3.4.0
-----
Development of the 3.4 series of MIRA concentrated on making assemblies with
30m to 100m reads more "liveable", i.e., reduce memory and disk footprint of
MIRA as well as improving run-times. At the same time, an updated assembly
strategy both for genome and EST / RNASeq data was devised to reduce the
influence of chimeras and intronic data on the assembly. Also MIRA is now
pretty smart in handling de-novo Solexa projects with "low coverage" (<30x) as
well as "high coverage" (>= 100x).

While we are at it: default parameters for Solexa de-novo were adapted to work
with at least 75mers. While doing de-novo assembly with smaller read lengths
is still possible for MIRA, the whole concept of ultra-short-read de-novo
assembly is a silly idea in the first place. So don't do it.

The new ability to handle IonTorrent data also made its appearance in MIRA as
implementing support for this kind of sequencing technology was comparatively
simple and straight forward. MIRA supports all kind of read lengths presently
on the market (100bp, 220bp) out of the box, but longer read lengths should
not pose a problem. Current IonTorrent data behaves very much like early 454
GS20 reads and I am curious whether Life will be able to perform the same
length and quality improvement within 12 month like 454 did in 2006. Time will
tell.

For PacBio, results are a mixed bag: CCS reads as well as error-corrected CLR
data works extremely well with MIRA, at least I'm happy how the E. coli
C227-11 demo data from the PacBio DevNet gets assembled. I suppose MIRA will
still need to get a couple more rules regarding the error profile of those
reads, but I'll be able to do that only once I've seen more data. What does
not work at all at the moment (and causes me some terrible headache) are the
CLR reads: those with an accuracy of only 80% to 85%. I'm not sure how to
tackle them efficiently.

For mapping assemblies, many smaller and bigger improvements ease the daily
life and improve results with those data sets. Exemplarily named should be
improved mapping quality of reads in highly repetitive regions of a genome
when the reference sequence is not optimal as well as the new ability to load
backbone sequences and annotation from GFF3 format files (saving will follow
shortly).

Quality control and automated clipping has been another focus in the past few
months. Notable developments there are automated clipping of known adaptors in
Solexa and IonTorrent data, improvements in the detection and avoidance of
chimeric reads and a some new automated editing algorithms which edit away
pretty clear cases of sequencing errors.

Regarding utilities, 'convert_project' has been revamped to be able to
convert large assembly or data files with less memory. It also got a number of
new options to get even more use cases covered. The new tool 'mirabait'
enables to quickly extract reads based on matching k-mers from a huge data
set.

For detailed changes, please consult the src/mira/CHANGES_old.txt file in the
source distribution.




3.4.rc5 (never officially released)
-----------------------------------
- changed default input type for 454 to FASTQ
- changed default -PF:qrml1:qrml2 for Solexa data to account for longer reads
- convert_project: -R now also works on single reads
- revamped IonTorrent guide
- revamped PacBio guide


3.4.rc4 (never officially released)
-----------------------------------
- bugfix mapping assemblies, again: some repetitive reads were still not
  optimally placed, new they really are.
- significantly faster file I/O on large files when a progress bar is
  displayed. This also improves I/O of other programs like convert_project,
  mirabait, etc.pp
- algorithm adaptation to enhance assembly of "long" IonTorrent sequences


3.4.rc3 (Release candidate 3)
-----------------------------
- bugfix mapping assemblies, again: some repetitive reads were still not
  optimally placed.
- bugfix convert_project, again.
- mapping assemblies: when reference sequence (backbone) loaded from GenBank
  or GFF3, tag MFSM recognised and used to forcefully merge mapped reads to
  coverage equivalent reads even if they contain mapping errors.


3.4.rc2 (Release candidate 2)
-----------------------------
- fixed major regression bug: some debug code went into 3.4rc1, slowing down
  most assemblies to a painful crawl.
- convert_project: fixed bug where contigs were always given default names
  (came in in 3.4rc1)
- convert_project: can now use -M and -s together
- convert_project: re-activated -d 


3.4.rc1 (Release candidate 1)
-----------------------------
- mapping of Solexa: MIRA now automatically keeps paired-end reads at contig
  ends unmerged to facilitate scaffolding. New parameter: -CO:msrkceu
- improved handling of repetitive stretches where the difference is a single
  indel.
- improved indel tagging in Solexa data.
- new parameter: --cwd or -DI:cwd to let MIRA change working directory
- new feature: on very long running assemblies with lots of passes (-AS:nop),
  one can now tell MIRA to terminate even if the required number of passes has
  not been reached yet by simply creating a file
    "<projectname>_assembly/<projectname>_d_chkpt/terminate"
  MIRA will then run 2 more passes and stop.
  Use case: people with complex assemblies who set -AS:nop to a high number
  and decide upon the intermediate results that the assembly is OK for them.
- reduced disk usage and less clutter in the tmp directory: a couple of log
  files is now written only when specifically requested (-MI:el).
- convert_project can now rename contigs/singlets via -R
- convert_project can now blind data (replace all bases by 'c') in reads, but
  also in contigs loaded from CAF and MAF. Use case: error reports.
- bugfix: some possible chimeras were sometimes both cut and extended into
  clipped area.
- bugfix: on some occasions, MIRA would stop while clipping Solexa or Ion
  Torrent adaptors
- bugfix: MIRA would very rarely stop an assembly when diagnosing a certain
  kind of impossible overlap. That has been changed, MIRA now puts a warning
  in the log but continues.


3.2.1.20
--------
- bugfix: chimera detection was not called anymore, bug entered in 3.2.1.18
- testing new pathfinder routines for better EST assembly, less prone to
  chimeras which passed clipping (for whatever reason).
- mate-pairs for overlaping reads now almost guaranteed to land in the same
  contig except in cases of major problems like chimeras etc.


3.2.1.19
--------
- the changes in 3.2.1.18 for low coverage Solexa assemblies sometimes led to
  explosion of memory needs, fixed.
- added parameter -AS:max_contigs_per_pass
- added capability to have MIRA just perform loading and clipping of data and
  then calculate hash statistics and read repeat information by seting
  -AS:nop=0
- bugfix: -FN:ssvsi:stvsi:smaltvectorscreenin were not parsed though they were
  documented to exist.


3.2.1.18
--------
- added new experimental option -SK:swcob to improve mapping accuracy of
  reads in repeats.
- assembly: some hash analysis done twice for no reason, fixed.
- assembly: better handling of "low coverage" Solexa assemblies
- mirabait: renamed parameters -l to -k and -k to -n
- mirabait: FASTQ loading now takes 33 as default to -o instead of 0
- mirabait: now contained in binary packages (symlink to convert_project)
- new extended log for railreads
- change: read ends now also get HAF tags even if these tags are < -SK:bph
  better overview in assemblies, but slightly more memory needs and larger
  CAF/MAF files
- new extended log: elog.dp.hashanalysis.lst
- bugfix: MIRA could sometimes abort the assembly in very high coverage
  situations. Fallout from change in allowed coverage to 1073741823 in
  3.2.1.16
- bugfix: SKIM missed borderline overlaps composed of only one hash
- bugfix: long standing bug in debug output of Read::AS_TEXT went to standard
  output instead of chosen stream or file


3.2.1.17
--------
- support for data from IonTorrent
- parameter -CL:cksa renamed to -CL:ckar and made it technology dependent
- bugfix in pathfinder where low coverage, small contigs with just a handfull
  reads might not have been built.
- bugfix pathfinder: some very short overlaps were sometimes not made when
  growing contigs
- bugfix: very short overlaps in fringe reads could lead MIRA to stop (bug
  triggered by new priotisation routines of 3.2.1.16)


3.2.1.16
--------
- less false positive repeat marker tags (mostly in 454 data due to identical
  454 reads with wrong PCR base inclusion)
- MIRA now automatically edits some conservative and unambiguous errors in
  alignments
- new overlap prioritisation routines reduce impact of PCR artefacts (mainly
  454, Solexa probably too)
- coverages for contigs can now reach 1073741823 instead of 16383 ... this
  should be really enough for any application in the foreseeable future.
  downside: memory needed for contig structures almost doubles.
- new parameter: -MI:somnrl to warn if long read names are encountered
- change: MIRA now handles CAF files with reads having defective clipping
  points more gracefully. Use case: deleting in gap4 consensus bases and at
  the same time deleting a read (to size 0)
- new documentation: MIRA utilities. Currently "convert_project" and
  "mira_bait"
- bugfix: -AS:urd now works again as advertised, but is switched off per
  default
- bugfix: in some cases the mapping mode could not start because MIRA created
  rail reads being too long 


3.2.1.15
--------
- fixed bug appeared in 3.2.1.13 where using chimera detection would lead MIRA
  to often abort the assembly


3.2.1.14
--------
- skim: changed ruleset to better characterise close repeats in mapping
  assemblies
- convert_project: new -t target "crlist"
- clipping: -CL:cbse:cksa were wrongly interpreted as technology dependend
  settings instead of common settings


3.2.1.13
--------
- new clipping routines / parametrisation for Solexa reads: -CL:cbse:cksa
- more efficient disk I/O for hash statistics analysis
- GFF parsing knows a couple more elements it previously did not know
- pathfinder now quicker at end of contig building for hybrid assemblies with
  Solexa data in hybrid assemblies.
- bugfix: repetitive reads could sometimes be mapped at wrong places (appeared
  in 3.2.1.9)


3.2.1.12
--------
- convert_project: new -s switch (splitting contigs into single files)
- bugfix: output of wiggle files could go into endless loop in very< rare
  cases
- bugfix: output of contigs as text is now well formated even for very long
  read names
- bugfix: bases per hash can now be 32 bases instead of 31 (or 16 instead of
  15 on machines which do not have 64 bit integer types)


3.2.1.11
--------
- bugfix for regression in 3.2.1.10: in mapping mode, SKIM made too many
  comparisons, in the order of a de-novo assembly
- convert_project: new -M switch (extracting reads out of assembled contigs)
- output of assembly as text: now correctly aligned for any length of read
  names
- bugfix output of wiggle files: illegal contigs with zero coverage do not
  lead to endless loop anymore
- small bugfix: maximum number of bases in -SK:bph is now 32, not 31.


3.2.1.10
--------
- MIRA has now clipping of Solexa adaptors (automatic only atm)
- bugfix SKIM: estimate of average hash frequence now more stable for
  borderline cases
- convert_project: new -N switch. Like -n, but output is sorted.
- convert_project: new -to target "cstats" for contigstats
- bugfix configure: now honours expat flags given to ./configure


3.2.1.9
-------
- SKIM now uses intermediate result purging, tremendously reducing size of
  result files in large assemblies.
- bugfix: SKIM sometimes saved wrong hits when there were effectively none
- bugfix: large increases of virtual memory needs between the passes have been
  cut down.
- convert_project: fixed segfault which was due to the new streaming method
- convert_project: fixed problem when saving contigs as FASTA (bugfix from
  3.2.1.8)
- convert_project: uses default '-t' the same as '-f' if '-t' is not given
- lowercase clipping: upper case mask character "X" does not break it anymore


3.2.1.8
-------
- fixed cause of segfault in 3.2.1.7 (came in from branch handling new
  consensus calculation)
- convert_project: added -C
- convert_project: less memory usage when source file does not contain contigs
- MAF loader: bugfix the progress bar (now gets correctly updated)
- SKIM: bugfix progressbar for large data sets. Now correctly shows progress
  instead of immediately show 100% and then seem dead.
- GFF parser: better checks and error messages
- GBF output: now writes ACCESSION entry


3.2.1.7
-------
Merged two dev branches: optimised SKIM evaluation and new contig consensus
calculations.
- new routines to choose and trim down SKIM hits. Save tremendous amount of
  disk and memory for high coverage projects (coverages >50x), especially when
  coverages reach >100x.
- new SKIM chooser also should lead to less misassemblies for those rare cases
  they still occure.
- fixed bug which made MIRA misestimating overlap scores in some cases.
- fixed bug which made MIRA misinterpret partial overlaps as complete
  overlaps.
- fixed memery allocation bug which led MIRA to perform a couple of more
  allocations than necessary.
- fixed possible division by 0 (entered in 3.2.1.6)
- temporary fix for error message "Asked for elements in start cache while not
  using startcache?" (introduced 3.2.1.5).
- convert_project: -n now also works with read collections, i.e., not only on
  contig names when loading contigs, but reads when loading data without
  contigs (like in all FASTA, FASTQ, but also CAF, MAF when no contigs are
  defined)


3.2.1.6
-------
- new parameters: -CL:mqtfer:mqtfernop:c3pp:c3ppmsl:c3ppmea:c3ppmgfe
- convert_project: new parameter -X
- mirabait: fixed bug which let mirabait not use single-hashes as bait
- fastaselect.tcl: fasta sequences now written out as single line
- fasta2frag.tcl: added -namesuffix and -startoffset


3.2.1.5
-------
- improvement: mapping assemblies which do not also build new contigs
  (-SB:abnc=no) now end much faster. E.g., 33 million Solexa reads in accurate
  mode take a day instead of almost 3 days.


3.2.1.4
-------
- improved internal handling of invalid overlaps. Effectively slashes memory
  needed there by a factor of 8 to 10. E.g.: in Solexa transcriptome 10m reads
  @ 100bp with quite some ploidy, uses just ~1 GiB instead of ~11.5 GiB
- MIRA now dumps out simple statistics on total bases and used bases in the
  output log.
- reactivated -n switch for convert_project


3.2.1.3
-------
- new program: mirabait
- bugfix: GenBank files with DOS newline characters now read correctly
- change: memory cap for the SKIM hit reduction routine: now 16 GiB instead of
  32.


3.2.1.2
-------
- backport changes from 3.3.9
- backbones can now be loaded in GFF3 format
- assembly of millions of EST/RNASeq now *way* faster. E.g.: 10m Solexa reads
  (100bp) down from several days to ~1 day. 
- fastqselect.tcl: bugfix for files having only "+" and not "+name..."
- fixed bug that could lead to segfault when not compiling with the bgtracking
  option.
- the 'mira' executables assumes 'mira' function if it is not named one of the
  recognised functions (mira, miraSearchESTSNPs, etc.pp)
- bugfix: -CO:fnicpst now working as advertised for Solexa
- change: 'N' at ends of reads now treates as masked per default
- fixed configure: --with-tcmalloc


3.3.9
-----
- migrated CVS to GIT


3.3.8
-----
- improved building on OSX, now should go through without problem
- improved F/R naming scheme not to trigger on names like BLABLA.234 (like
  projects stored at the NCBI SRA). Important for projects with paired-end
  data.
- backbone reads loaded from FASTA and GBF as well as backbones from CAF and
  MAF which do not contain contigs now don't have a read naming scheme,
  suppressing unnecessary error messages if the names contain "."
- for mapping Solexa data, MIRA now reduces unnecessary mapping attempts when
  no more reads match a contig. This should save a bit of time, especially in
  projects with lots (>1000) of reference sequences.


3.3.7
-----
- improved OS X compatibility of the build script which should now be able to
  build static MIRA binaries for OS X versions >= 10.5
- maybe improved BOOST recognition on some systems due to updated autoconf
  macros


3.2.1
-----
- while reading NCBI TRACEINFO XML, MIRA now complains if it sees either
  insert_size or insert_stdev but not both
- -AS:mrpc could lead to non-sequential contig numbering, fixed
- change: Sanger naming scheme now expects only a postfix length of at least 3
  instead of 4
- error message when reads too long are encountered now clarifies size of
  MAXREADSIZEALLOWED
- change: MIRA now checks read names for funny or illegal characters
- bugfix when loading FASTA files: \r on name lines in DOS/Windows files was
  not handled properly
- bugfix in fasta2frag.tcl: when qualities were available at loading, only
  zeros were written to fragment quality file. Bug entered sometimes in the
  3.1.x series.


3.2.1rc2 (3.3.6)
----------------
- fixed bug introduced in 321rc1 which led to more than necessary disk
  operation in SKIM hit reduction after the first pass of MIRA.


3.2.1rc1 (3.3.5)
----------------
- "--job=mapping,solexa" did not set really good values for -SK:mhpr, leading to
  sometimes sub-optimal mappings. New value should be good for all but the
  most repetitive eukaryotes.
- "--job=mapping" now uses "-AS:nop=1:rbl=1" to conform more to user
  expectations of a mapping.
- test: reducing SKIM hits now uses 30 GiB at most for array, regardless of
  memory settings. Furthermore, array is re-used between passes. Should reduce
  MIRA memory footprint.
- bugfix: when Solexa sequences in FASTA and FASTA quality files were not in
  the same order, wrong assignmesnt of qualities could happen under certain
  circumstances.


3.3.4
-----
- performance bugfix: SKIM reduction in mapping projects with several millions
  reads now runs considerably faster (e.g. seconds instead of 45 minutes when
  using 15 million reads)
- the SSAHA2 screen reading routines now also read SMALT output files where
  the "-f ssaha" output option was used
- small enhancement: no second round of proposed end clipping if first round
  did not yield any clip.


3.3.3
-----
- bugfix: mapping with >= 8 strains is now possible again for non-Solexa reads
- enhancement: mapping with >= 8 total strains and <= 7 Strains with Solexa
  data is now possible
- silenced error output for 'stat' calls on BSD and BSD-like systems
- reworked *_info_assembly.txt file to be a bit clearer
- reduced compiler warnings (replaced deprecated gcc hash_map with BOOST
  unordered_map)


3.3.2
-----
- bugfix: miraSearchESTSNPs failed to save correct data for step3. Error was
  introduced somewhere between 3.0.5 and 3.2.0
- expanded new pathfinder functionality which caches start sites also to EST
  assemblies
- added -MI:lcs:lcs4s
- compiling: building MIRA from a source package now works using calls like
  "../some/other/dir/configure"
- compiling: migrated 12 year old home-brew solution for including flex files
  to now standard makefile functionality.


3.3.1
-----
- saving skim hits in binary instead ASCII saves between 20 and 60% disk space
  for skim hits. Also improves NFS performance a bit due to added
  buffering. Also improves multithreading performance: now at almost n*100%
  where n is number of threads. Lastly, improves performance of SKIM reduction
  filtering algorithm.
- updated docs to reflect interferences between -AS:mrpc and -OUT:sssip:stsip
- chapters in the manual now have readable (and constant) reference names
- To help 90% of the users doing a mapping assembly: the default name for
  backbones / reference sequences is now "ReferenceStrain" and not an empty
  string (-SB:bsn). Also, the default name for reads without strains is now
  "StrainX". New parameters -SB:ads:dsn.
- testing new pathfinder functionality which caches start sites (-PF:mrcbs)
- added missing -OUT:orm:otm
- fixed bug which could lead to an out-of-memory error when no overlaps were
  present in an assembly
- removed deprecated -GE:kcim parameter
- changed behaviour: in EST assemblies, statistics for large contigs are
  computed for contigs >= 1000 bases instead of 5000 as for genomes


3.2.0
-----
Changes since 3.0.5:

- Support for data from Pacific Biosciences.

  Though the 3.0.x (and even earlier 2.9.x) versions could handle PacBio data
  when configured accordingly, MIRA 3.2.0rc1 now officially supports PacBio.

  Beside the usual support for non-paired and "paired-end" data, MIRA also
  has a new automatic editor for PacBio reads which should be useful when
  dealing with "elastic dark inserts" (longer stretches of unread bases whose
  length is only approximately known) in PacBio strobed sequencing mode.

  With this, MIRA should be able to deal with strobes of unread bases up to
  ~400 bases without having to split strobed sequences in multiple read-
  pairs. Simulation with bacterial data show that PacBio strobe reads in
  200/200 mode (sequence 200 bases, skip 200 bases, repeat) and having 3000
  sequenced bases can reconstruct a bacterial genome quite well (1 contig,
  correct genome organisation), leaving only "clean up" work of getting some
  bases right via, e.g., hybrid approach by complementing with e.g. Solexa
  (Illumina) data.

  Though I admit that PacBio support is a long shot (I don't have real PacBio
  data at the moment), I expect MIRA to be "good enough" for first test with
  real data (for those people out there actually having access to some). On
  the other hand, I haven't seen any other assembler yet being able to
  support strobed reads without splitting them.

- Fully revamped manual

   In an exercise of self-defence (too many mails in my inbox), I've updated
   the manuals to DocBook format and considerably expanded them. The result:
   nicer manuals in HTML and PDF format, with extensive walkthroughs There are
   also screenshots in colour. And variegated.

- The "usual" improvements and bug-fixes

  If PacBio and the revamped manuals had not been on the schedule, the
  "usual" improvements would have a had been described more prominently in
  this announcement. Things being what they are, I'll just mention 
    - hybrid Sanger/Solexa or 454/Solexa of bacteria now finish within hours
      instead of days.
    - longer contigs
    - less memory utilisation (thanks to Google TCMalloc library)
    - better support for SSDs
    - warnings when using NFS


The MIRA 3 versions are the result of a long development to make assembly
of Sanger, 454 and Solexa (Illumina) data as easy and straightforward as
possible while keeping a maximum accuracy.

Another focus was to make it possible to use results from Solexa mapping
projects in current finishing programs, not only viewers. MIRA introduces for
that the notion of coverage equivalent reads (CER) which reduces the data
volume by 70 to 90%. This allows painless use of of such data sets in gap4 and
consed.

A lot has changed since the 2.8.x series of MIRA, the following list has just
a few highlights which came in during the 2.9.x series:

- sequencing technologies: MIRA handles different sequencing technologies
  independently from each other and has specialised routine for working with each
  of them.
- command-line parameters: MIRA has now a handful of "Do-What-I-Mean" one-stop
  switches which allows to configure the assembler for 90% of all use
  cases. Furthermore, many parameters can be adjusted for each sequencing
  technology so that the assembly engine can be tweaked for very specialised
  cases if needed.
- all sequencing technologies (Sanger, 454, Solexa) have now
  - recognition of chimeras
  - new assembly routines to for improved repeat resolving
  - improved data preprocessing that gets rid of low quality data and
    sequencing errors at ends of reads ... even when no quality data is
    available.
- 454 data:
  - fully developed capability for de-novo and mapping assembly of 454 data
    (paired and non-paired)
  - automated contig editor to remove most obvious and/or annoying sequencing
    errors
  - improved consensus calling streamlined to minimise the dreaded homopolymer
    problem
- Solexa data
  - can handle Solexa data of any length, no restriction to very short
    sequences.
  - memory/space saving: MIRA has special mapping mode which creates data
    so that widely used finishing tools like gap4 and consed can load these
    projects and still be fairly quick
- alignments enriched with features: MIRA adds information like repetitiveness
  or repeat marker bases as tags in the assembly so that these can be used
  during finishing
- assembly information files: MIRA writes more information files which can be
  easily parsed and/or read
- mapping assemblies: MIRA has a full SNP analysis for prokaryotic data
- comprehensive tables and HTML result files (mapping assembly): the
  convert_project program can now create easy to use tables and HTML files
  which show the data in a way suited for less computer-interested people
  (biologists etc. :-)
- memory management: MIRA can now be told to use an upper limit of RAM.
- file formats: MIRA can now parse or write more file formats. Notable are
  change from SSAHA to SSAHA2 for clipping, FASTQ for data input and MAF
  format for output.
- MacOS support: MIRA now compiles on MacOS-X
- speed and memory: compared to 2.8.x, MIRA now uses way less memory and is a
  lot faster.
- tons of other features, tweaks and bug fixes. See the CHANGES_old.txt file


3.2.0rc3a
---------
- fixed error: setting -MI:sonfs from the command line did not work
- added code to catch erroneous "-- something" in parameters and give an
  appropriate error message
- PacBio now does not expect XML data as default


3.2.0rc3
--------
- regression problem: changes for convert_project led to incredibly long running
  times for computing consensus in mapping assemblies. Fixed.
- fixed configuration error for miraSearchESTSNPs: using --job=esps2 did not
  load sequences from step1
- new parameter -SK:acrc for switching on/off searching of reverse complement
  hits


3.2.0rc2
--------
- new parameter category "MISC" with first new parameter "stop_on_nfs"
  (-MI:sonfs)
- added -DI:lrt to be able to put the log directory to other locations (i.e.,
  support SSDs or take the heat off of NFS mounts)
- added -CL:pechsgp to be able to switch handling of Solexa GGCxG problem if
  wanted.
- larger work on convert_project functionality which should now behave a bit
  more as user might expect (see below)
- when used on an assembly file with multiple strains (caf or maf),
  "convert_project -t fasta" now also creates files with combined output of
  all strains
- reversed default and meaning of '-u' in convert_project: per default filling
  of strain is off, can be switched on
- '-u' in convert_project now also fills the '@' "base" (which stands for "no
  coverage by this strain")
- '-v' in convert_project now works "per strain" and not "per total coverage"
- added check for zlib in "./configure"
- changed behaviour: -SK:mnr is now switched on by default for EST projects
- changed behaviour: the *nastyseq* file in the log directory has been upgraded
  to an "info" file, i.e., it is now in the info directory. Furthermore, it
  can log (on demand) not only sequence parts covered by MNRr tags, but also
  HAF5, HAF6 and HAF7. New parameter: -SK:rliif
- bugfix: -OUT:rrol did not work, old logs were not removed (bug introduced in
  3.1.7)
- updated docs. Section on "things you should not do", added description of a
  couple of parameters which did not make it to documentation yet, etc.
- HTML documentation lacked underline-emphasis, fixed.


3.2.0rc1
--------
- completed change to new DocBook manual system, revamped and extended
  manuals.


3.1.16
------
- change: for genome assemblies, MIRA now builds longer contigs when large
  paired-end libraries are present (10k, 20k or more)
- change: filter for bad Solexa reads now writes names to clipping log instead
  to standard output.
- fasta2frag.tcl: added -P, changed -r to work also on paired-ends
- bugfix: PacBio data without elastic dark inserts led to segfault
- bugfix: -SK:mnr was always treated as "yes", even if set to "no".
- bugfix: fixed miraSearchESTSNPs stopping in step2 if step1 result files were
  empty.
- bugfix: in 3.1.15, the logic to automatically set -CO:emea=1 worked only for
  Sanger reads, now for all reads.
- bugfix: read tag comments in ACE file missed a newline


3.1.15
------
- new parameter -CO:emeas1clpec. Automatically sets emea to 1 if proposed end
  clipping is used (ends will be "clean"). Improves recognition of
  misassemblies in cases where only the outer fringes of reads differ.
- change in template handling: to be lenient, MIRA internally added/subtracted
  10% of the given insertsizes (or at least 1kb). Not anymore! This would give
  problems with very small libraries (Solexa) or when the given values were
  "lenient enough" and were made "too lenient" by this and subsequently
  flagged in different post-processing tools.
- change in handling template insert size info from XML: previously, MIRA set
  stdev to a minimum of 500 bases and used 2*stdev to calculate minimum and
  maximum insert sizes. The 500 bases minimum rule has been removed, and now
  using 3*stdev
- new parameter: -GE:tpbd to give template partner build direction on the
  command line. Defines whether the template partner of a read (in a
  read-pair) must have the same direction (1) or reverse direction (-1) in a
  contig.
- change: when --job=...,454 is used, the default minimum overlap is not 40
  anymore, but 20. 40 was too conservative, overlaps at weak contig joins were
  discarded too often.
- improved graph reduction algorithm: some more small overlaps at low coverage
  sites are taken to Smith-Waterman. This helps to find some more weak contig
  joins.


3.1.14
------
- speed up of routine to find and mark IUPAC bases and unsure bases (IUPc &
  UNSc). Very noticeable when using annotated genomes as mapping reference.
- bugfix: IUPC & UNSc were not searched for anymore (introduced in 3.1.12 with
  the -CO:asir bugfix)
- re-activated '-d' in convert_project
- adjusted miramem estimator for mapping of Solexa reads


3.1.13
------
- improvements for large assemblies with millions of reads where setting up
  data for new contigs during build is sped up. Especially noticeable in EST
  assemblies, but also genome assemblies with Solexa.


3.1.12
------
- new option to speed up assemblies with millions of reads: -AS:mrpc controls
  the minimum number of reads a contig must potentially have before it is
  really assembled. This prevents all the small junk contigs with very low
  numbers of reads in, e.g., Solexa sequencing to be assembled and can speed
  up the assembly by days.
- MIRA now uses the tcmalloc library from Google perftools if available. It is
  highly recommended as it optimises memory allocation and saves a lot of
  memory on multiple pass assemblies. E.g., memory usage for 810k 454 FLX
  reads, 45x coverage, 5 pass genome de-novo accurate:
              3.0.5    8272988 kB
             3.1.11    8273012 kB
             3.1.12    9492956 kB
     3.1.12tcmalloc    6758916 kB
- change: adapted some estimators in miramem, hopefully giving better
  estimates for RAM usage during MIRA assemblies.
- bugfix: array iterator overrun in contig building which had probably no
  noticeable effect. If, then perhaps rejecting weak matches it would have
  barely accepted.
- bugfix: -CO:asir sometimes set repeat markers instead of SNP markers.
- bugfix: mira could try to check physical presence of SCF data even for
  non-Sanger reads


3.1.11
------
- optimisation: memory pre-allocation routines for read growth help to get
  down memory fragmentation and hence less memory requirement
  overall. Especially noticeable in high coverage 454 sequencing or with
  strobed PacBio reads.
- bugfix: -CO:mr=no was not fully respected. While not used during contig
  building, possible repeats were always marked in result files and then
  tranferred to following iterations.
- bugfix extendADS(): acquireSequences() could throw due to 0 length of a
  sequence


3.1.9
-----
- change: mira will stop immediately if it is launched with parameters that
  suggest miraSearchESTSNPs should be used instead
- bugfix: est assembly used genomic pathfinding routines instead of EST
  routines, leading to more contigs with almost identical consensus.
- bugfix: miraSearchESTSNPs pipeline for steps 2 and 3 did not load results
  from previous steps.
- bugfix: fastqselect.tcl script printed out the name of the firt read twice


3.1.8
-----
- changed wiggle file: more info on strains in the description and smaller
  files, using a stepping and span of 4 instead of 1
- new script: fastqselect.tcl (like fastaselect.tcl, but works on fastq)
- new 3rd party scripts by Tony Travis (qual2ball) and Lionel Guy
  (caf2aceMiraConsed.pl) to simplify integration of MIRA assemblies in
  Consed.
- updated the 3rd party documentation "Instructions for scaffolding MIRA
  contigs using paired ends" from Gregory Harhay


3.1.7
-----
- changed method to remove old files which hopefully minimises the number of
  files which fail to be removed during a run
- test changes to adapt to >= 2^32 skim hits
- adapted post-SW scoring for PacBio
- fixed bug: array underrun in alignment code, introduced in 3.1.3 (don't I
  love valgrind :-)


3.1.6
-----
- changed automatic memory management to use all memory minus 15% instead of
  of minus 10%.
- speedup of SKIM when running in multiple threads: removed unecessary call of
  a mutex lock (leftover from debugging code). Very noticable when running at
  higher thread numbers.
- bugfix: race condition in SKIM leading to wrong assessment of memory needs
- bugfix: calculation for assessment of memory needs was faulty, leading to
  similar problems as the race condition in SKIM above
- workaround: mapping with data containing artificial reads with lengths of
  several kilobases led to too high values for rail read length being
  computed. Fixed by capping at 18kb.
- temporarily switched off skim junk detection, it might be faulty at high
  coverages


3.1.5
-----
- bugfix: reading SSAHA2 data gave an error for Solexa reads beginning with a
  'N' (now really)
- bugfix: some SSAHA2 input files led to infinite loops
- calculation of SW alignment score sped up very slightly


3.1.4
-----
- bugfix: when read extension was used for any sequencing technology, it was
  also applied for reads of technologies where read extension was not wished.
- fixed compilation: new use of the stringcontainer could lead to a static
  initialisation fiasco (dependent on linker used at compile time) and
  subsequent crashes directly after start.


3.1.3
-----
- added first support for PacBio
- fasta2frag.tcl gets a mode to simulate strobed data
- reduced hits reported by SKIM when a reads fully covers another. Especially
  useful for hybrid assemblies of short / long reads.
- slight improvement of SW parametrisation and alignment algorithm (for
  strobed reads)
- fixed error with read names when using mapping mode
- fixed potential unwanted increase in memory consumption while loading SKIM
  hits.
- fixed compile problem of ./src/caf/caf_flexer.flex on CentOS
- fixed small bug in ./configure.in where rescue values for BOOST paths were
  not set correctly for some systems


3.0.1 (backported from 3.1.2)
-----------------------------
This version fixes a few quirks and problems of the initial 3.0.0
release, some of them leading to MIRA aborting or even hard crashes.

MIRA also was a bit too picky in 3.0.0 for joining some reads. Due to changes
and algorithm optimisation, there should be notable improvements in contig
lengths (N50 etc.) in genome assemblies with bad data. In EST assemblies,
chimera cutbacks are now disabled by default, leading to less cutbacks.

Important note for users of sff_extract in paired-end projects: please switch
to the newest version of sff_extract (>= 0.2.8) as the old ones contain a bug
and do not reverse quality values for reverse reads.

- changed SSAHA2 parser to allow for pathological case of empty vector names
- changed method for average coverage estimation slightly to better cope with
  extremely skewed distributions (seen in some EST data)
- added workaround to allow usage of SSAHA2 screening data with Solexa reads
- improved speed of pathfinder algorithm for repetitive 454 reads
- improved concurrency of SKIM output, better use of available thread capacity
- added method to propose smaller cuts at the end of reads in SKIM (-SK:asjdc)
- added flags to control chimera cutbacks (-SK:ascdc) and junk cutbacks 
  (-SK:asjdc). On by default for genome assemblies, off for EST assemblies
- increased speed of SKIM hit reduction for assemblies with long and short
  reads (Solexa & ...)
- improved handling of reads with problematic ends which could lead to
  premature stop of contig building
- reduced memory need for internal read structure. As part of this and only
  user visible effect (if at all), the Staden-ID of CAF files is not supported
  anymore
- reduced memory needs for tags. Side-effect: slight speed improvement of
  algorithms using tags
- bugfix: consensus of Solexa bases only with N now results in N instead of a
  space
- bugfix: FASTA file with multiple equal read names now lead to MIRA stopping
- fixed critical buffer overruns that could lead to weird errors or even to
  MIRA crashing hard with segmentation faults
- fixed the annoying "len1 or len2 == 0 ?" bug (turned out to be side-effect
  of chimera clipping)
- fixed error in SKIM parametrisation which could lead in some cases to long
  run times, excessive memory consumption and data corruption.


3.0.0
-----

The MIRA 3 versions are the result of a long development to make assembly
of Sanger, 454 and Solexa (Illumina) data as easy and straightforward as
possible while keeping a maximum accuracy.

Another focus was to make it possible to use results from Solexa mapping
projects in current finishing programs, not only viewers. MIRA introduces for
that the notion of coverage equivalent reads (CER) which reduces the data
volume by 70 to 90%. This allows painless use of of such data sets in gap4 and
consed.

A lot has changed since the 2.8.x series of MIRA, the following list has just
a few highlights which came in during the 2.9.x series:

- sequencing technologies: MIRA handles different sequencing technologies
  independently from each other and has specialised routine for working with each
  of them.
- command-line parameters: MIRA has now a handful of "Do-What-I-Mean" one-stop
  switches which allows to configure the assembler for 90% of all use
  cases. Furthermore, many parameters can be adjusted for each sequencing
  technology so that the assembly engine can be tweaked for very specialised
  cases if needed.
- all sequencing technologies (Sanger, 454, Solexa) have now
  - recognition of chimeras
  - new assembly routines to for improved repeat resolving
  - improved data preprocessing that gets rid of low quality data and
    sequencing errors at ends of reads ... even when no quality data is
    available.
- 454 data:
  - fully developed capability for de-novo and mapping assembly of 454 data
    (paired and non-paired)
  - automated contig editor to remove most obvious and/or annoying sequencing
    errors
  - improved consensus calling streamlined to minimise the dreaded homopolymer
    problem
- Solexa data
  - can handle Solexa data of any length, no restriction to very short
    sequences.
  - memory/space saving: MIRA has special mapping mode which creates data
    so that widely used finishing tools like gap4 and consed can load these
    projects and still be fairly quick
- alignments enriched with features: MIRA adds information like repetitiveness
  or repeat marker bases as tags in the assembly so that these can be used
  during finishing
- assembly information files: MIRA writes more information files which can be
  easily parsed and/or read
- mapping assemblies: MIRA has a full SNP analysis for prokaryotic data
- comprehensive tables and HTML result files (mapping assembly): the
  convert_project program can now create easy to use tables and HTML files
  which show the data in a way suited for less computer-interested people
  (biologists etc. :-)
- memory management: MIRA can now be told to use an upper limit of RAM.
- file formats: MIRA can now parse or write more file formats. Notable are
  change from SSAHA to SSAHA2 for clipping, FASTQ for data input and MAF
  format for output.
- MacOS support: MIRA now compiles on MacOS-X
- speed and memory: compared to 2.8.x, MIRA now uses way less memory and is a
  lot faster.
- tons of other features, tweaks and bug fixes. See the CHANGES_old.txt file


2.9.59 (3.0.0)
--------------
- moved all output directories into one directory named <projectname>_assembly
- added 3rd party documentation to distribution packages
- mapping 454 reads in 'accurate' mode now does not automatically switch on
  the feature to also build new contigs (which comes rather unexpected for
  users and also completely changes runtime behaviour)
- base jiggling in homopolymers should be further reduced (problem of 454
  data)
- added FASTQ as conversion target to convert_project
- added quickswitch -noquality
- renamed -LR:llc to -LR:lcc (corrected typo)
- clip lowercases (-LR:lcc) now does not clip if all sequences to be clipped
  contain just lower case
- SNP evaluation routines now handle feature rich GenBank files more
  gracefully.
- streamlined tagging: in mapping assemblies, positions having received SNP
  consensus tags now don't have less important tags denoting problematic
  positions (UNSc, IUPc etc.)
- improved alignment and consensus calling routines reduces homopolymer errors
  in 454 data by ~30% - 60%
- clearer error message from MIRA when FASTQ file cannot be loaded
- changed behaviour of -CL:msvs to now load SSAHA2 data instead of SSAHA data
- changed behaviour: MIRA now stops if it encounters input reads with no names
- changed behaviour: backbone reads in mapping assemblies now count as normal
  coverage.
- changed behaviour: miraMem now calculates with Titanium average length of
  400 instead of 475
- changed behaviour: loading of Solexa data now defaults to "new" phred style
  qualities (-LR:ssiqf=no)
- changed behaviour: for Solexa data, MIRA now calls deletions more
  aggressively when in a tie with another base.
- changed behaviour: for Solexa data, MIRA now calls less IUPAC bases.
- changed behaviour: the HAsh Frequency (HAF) tags are now set slightly
  differently to show the potentially dangerous sites. Furthermore, HAF6 is
  now defined as 'heavy' repeat, HAF7 as 'crazy'
- changed naming of contigs with only repetitive sequence: now named *_rep_c
  instead of *_lrc
- fixed bug: automatically calculated values for -SB:bro and -SB:brl were a
  factor 2 too low.
- fixed bug: de-novo assembly with more than 8 strains led to MIRA stopping
  where it should not.
- fixed bug: log file "Edit.log" from the Sanger EdIt now created only when
  really needed
- fixed bug: --job=esps2 and esps3 failed to start because of flawed
  default parameters
- fixed bug: convert_project did not properly convert singlets into EXP files
  ... the directory was not created.
- fixed bug: convert_project -m replaced qualities of reads with '30' even if
  those had been present in the input.
- fixed automatic recognition of Sanger FASTQ format where some data sets led
  to slightly offset quality values
- fixed bug when pre-loading EXP files that contain quality values (thanks to  
  David Phillip Judge for mailing the bug ... AND the fix)
- fixed small bug in loading EXP files with certain entries which went
  unnoticed for 11 years *sigh*
- fixed small bug in CAF and MAF output when gaps are present at the 3' end of
  reads.
- fixed bug that could lead to segmentation faults when calculating assembly
  statistics info (thank you Valgrind)
- fixed bug in how MIRA checked whether it runs in 32 or 64 bit.
- fixed Trac bug #5 (sometimes error in transferContigReadTagsToReadpool())
- fixed bug: in mapping assemblies, not all possible alignments of repetitive
  parts reads where given back by SKIM
- fixed bug: in rare cases (mostly projects without templates or paired-end),
  MIRA joined repeats where it should not
- fixed bug: in mapping assemblies with the option to build new contigs, MIRA
  often preferred to make new contigs from reads that had a difference to the
  reference sequence
- fixed inconsistency in calculation of N50/N90/N95


2.9.58 (aka V3rc4)
------------------
- MIRA now also compiles again on Apple Mac OSX (yay!)
- scripts added to distribution: fixACE4consed.tcl, fastaselect.tcl and
  fasta2frag.tcl
- new clipping: clip lowercases (-LR:llc). Made for 454 data, but can be used
  with other sequencing types, too.
- MIRA now switches off automatic memory management when system information
  cannot be gathered
- change to get SKIM working on 31 base hashes also on 32 bit machines where
  the compiler knows 64 bit data types
- slightly changed messages for STDOUT log when writing cluster info to disk
- added option "-r f" to convert_project
- convert_project can now convert from and to MAF
- changed -GE:gbmf to -GE:kpmf (sorry for the inconvenience)
- miraSearchESTSNPs does not load "me_stepX.par" files anymore (it parses the
  command line like normal mira now)
- default parameters for miraSearchESTSNPs have been chaged in step 1 and
  2. Notably, the automatic editor is switched of in step 1, but switched on
  in step 2. Furthermore, both steps change from single pass/multi-loop
  (-AS:nop=1:rbl=...) to multipass/multi-loop (-AS:nop=...:rbl=...) setup
- fixed bug in automatic memory management where RAM allocated was actually
  less than minimum asked for and could lead to disastrous assembly results
- fixed bug that prevented recalculation of contig consensus when loading CAF
- fixed bug in contig building where in rare cases a missed alignement led
  to MIRA stopping
- fixed bug that led to extremely long run times and suboptimal contigs in
  rare cases
- fixed bug that led to MIRA stopping in rare cases during mapping of Solexa
  reads
- fixed bug that led to MIRA missing internal functionality of merging short
  reads after loading contigs from CAF or MAF
- fixed segmentation fault in parsing the "--job" parameter when only "est"
  was given
- fixed bug that led MIRA want to load a file named ".maf" when trying to load
  a CAF format.
- fixed bug where loading of sequences in CAF failed as MIRA thought there was
  no data in the file
- added logfile to replay eventual errors during contig building


2.9.57 (aka V3rc3)
------------------
- added -GE:amm:gbfm:mps as first trials for automatic memory management
- added -FN:bbin for naming backbone input files
- added -LR:wqf and -AS:epoq for more thorough checking of presence of quality
  values, by default MIRA now stops if quality files are expected but not
  found or when reads with no quality values are present in the assembly.
- parameter parsing now checks that parameters that are specific for sequecing
  types are in a correct section (SANGER_SETTINGS, 454_SETTINGS, etc.pp) and
  that common parameters are in a COMMON_SETTINGS section. SANGER_SETTINGS is
  now no longer an alias to COMMON_SETTINGS. Updated documentation to reflect
  changes.
- reduced number of memory allocations for Smith-Waterman alignments
- estimator for internal memory usage got better
- fixed bug in wrong parameter combination for --job=est assemblies  
- fixed bug in contig building routine that sometimes stumbled over IUPACs in
  mapping mode ("logical error 2")


2.9.56 (aka V3rc2hf3)
---------------------
- reduced memory needs of SKIM results for projects containing lots of small
  reads combined with lots larger reads (i.e. Solexa + Sanger and/or 454)
- fixed small bug in logfile tracking that stored too many copies of given
  logfile names and unnecessarily gobbled up hundreds of megabytes of RAM for
  projects with large numbers of contigs (*sigh*)
- (for compiling) configure script now checks presence of expat


2.9.55 (aka V3rc2hf2)
---------------------
- (for compiling) tweaked configure script to better handle cases with
  programs/include files installed in non-standard locations
- (for compiling) configure script now checks presence of flex++


2.9.54 (aka V3rc2hf1)
---------------------
- changed loading of CAF files to behave more like other file types
- fixed floating point error in miramem
- (for compiling) removed unnecessary (and sometimes counterproductive) .flex.C
  files from source distribution.


2.9.53 (aka V3rc2)
------------------
- fixed bug in ACE output of consensus tags (had "C}}" instead of correct
  "C}\n}" for closing tags)
- fixed another bug in ACE output, this time read tags (forward tags were
  sometimes written out in reverse)
- re-enabled output of results as HTML. Not ideal, but works.
- re-enabled "-t html" in convert_project
- due to introduction of FASTQ as input format, the abbreviation switch for
  naming FASTA quality input (-FI:fqi) needs to be renamed to
  "-FI:fqui" as "-FI:fqi" now names FASTQ input files.
- reduction of memory usage for cases where the possible_vector_clip is not
  needed (memory is not allocated, default for all assemblies without Sanger
  data)


2.9.52 (aka V3rc1b2)
--------------------
- implemented aggressive memory saving for Solexa reads which reduces the data
  stored per read base from 9 bytes to 5 bytes (45% reduction). The downside:
  the result files which have alignment positions (CAF, ACE etc.) do not show
  insertions and deletions anymore in the coordinates. I.e., edits cannot be
  traced back.
  As no other assembler has this info and no finishing program I know uses
  this info anyway, I guess this is ok.
- optimisation: MIRA now uses less memory constructing coverage equivalent
  reads (mapping assemblies -CO:msr=yes)
- bugfix: on large contigs with lots of reads, MIRA now uses significantly
  less temporary memory for all mapping assemblies (e.g. 3GB less when mapping
  6m Solexas)
- compile instructions for NetBSD (courtesy of Thomas Vaughan)
- re-activated and upgraded TCS output (transposed contig summary)
- implemented "-a" parameter for convert_project
- wrote documentation for the changes


2.9.51 (aka V3rc1b1)
--------------------
- reduced overhead of reads by 32 bytes (on 64 bit architectures). That's
  320Mb for 10m reads :-)
- testing majority vote of 66% for gaps in consensus calling of 454 reads.
  (NOTE: that code is is faulty as it always evaluates to true and sometimes
  leads even to a division by zero error. Just disregard this version.)


2.9.50 (aka V3rc1b)
-------------------
- bugfix: mixed FASTA and CAF loading works again
- bugfix: "--<filetype>" now sets file type for all sequencing technologies as
  intended.


2.9.49 (aka V3rc1a)
-------------------
- bugfix: CAF loading works again
- new tag: DGPc (Dubious Gap Position on Consensus). Set when the number of
  gaps at a consensus position is between 40% and 60% of the next most
  frequent base. E.g.: A/* = 10/7. But also when A/C/* = 9/10/7


2.9.48 (aka V3rc1)
------------------
- bugfix for hybrid assemblies involving Solexa reads: some Solexa reads did
  not make it into contigs
- optimised SKIM reduction routine for hybrid assemblies involving Solexa: the
  overlap graph generated uses less memory.
- optimised SKIM/Pathfinder interaction for short Solexa reads (<60 bases,
  longer reads were not that much of a problem) which allows better de novo
  with Solexa.
- major speed increase in pathfinder module for large de-novo assemblies with
  millions of reads. E.g. 2 pass de-novo with 6m Solexa paired-end 36mers goes
  down from 4 hours 20 minutes to ~2 hours.
- major speed increase for hybrid assemblies involving Solexas in pathfinder
  module: a 5 pass de-novo with 800k 454 reads and 3.3m Solexas goes down from
  >1.5 days to 12hrs
- reduction of memory needs for Solexa data (e.g. ~1.2GB for 7.3m Solexa
  40mers). (There was an ommision of container capacity reservation since
  2.9.44x1)
- convenience change: -SB:brl:bro can now be set to 0 for automatic
  determination of optimal values by MIRA in mapping assemblies (now
  default).
- convenience change: -AL:shme can now be set to 0 for automatic determination
  of reasonable value (now default)
- change for genome assemblies via quick switch: masking of nasty repeats is
  now turned on, copy threshold at 100x expected frequency
- change for genome assemblies with 454 data via quick switch: proposed end
  clips now more stringent and enforces 27 instead of 17 bases clear space


2.9.47
------
- small change in default parameters when using Solexa data (alone or hybrid)
  to better adapt to larger difference to reference (mapping) or low coverage
  (mapping and de-novo).
- reworked small Solexa examples in minidemo directory (mapping and denovo)


2.9.46x3
--------
- unified parameters for loading different sequencing technologies: added
  --fastq, -LR:lsd:ft:fqqo and -FI:fqi; removed
  -LR:lsand:l454d:lsxad:lsidd:sanft
- MIRA can now load data FASTQ format, routines are courtesy of Heng Li at the
  Sanger Centre. For Solexa data there's also an automatic recognition of
  whether it's in Sanger, Solexa 1.0 or Solexa 1.3 format.
- convert_project can subsequently also load FASTQ and gets an additional -o
  parameter.
- debris file does not contain same read name multiple times
- for build process: moved check for isblank(3) to configure.in
- new requirement for compiling: zlib
- MIRA now gets compiled with -O3 by default on most platforms
- further changes to configure script to allow correct compilation of 64 bit
  on platforms that compile with 32 per default 
- MIRA confirmed to compile on OpenSolaris with BOOST (yay!)


2.9.46x2
--------
- some more tweaks to ./configure (better lex/flex handling, better expat
  recognition, more chatty in case of boost problems)


2.9.46x1
--------
- just some tweaks in the build process to fix reported problems during
  compilation or linking


2.9.46
------
- updated major parts of the documentation for anything related to using
  Solexa reads
- new help file on how to assemble 'hard' genomes (mostly geared towards
  eukaryotes, but some prokaryotes also have a tendency to be nasty)


2.9.45x4
--------
- removed testcode that limited coverage to 5000x, now back to theoretical
  limit of 16383x.
- reactivated "html" as convert option for convert_project.
- activated "asnp" and "hsnp" as convert option for convert_project.
- removed dumping of debug information when using -t asnp or hsnp in
  convert_project


2.9.45x3
--------
- tuning of repeat detection for Solexa reads leading to less false positive
  repeat markers for typical Solexa miscalls
- tuning of base calling for Solexa reads leading to less IUPAC bases at
  places with typical Solexa miscalls


2.9.45x2
--------
- Probable fix for compiling src/mira/dataprocessing.C on Red Hat systems
  (inclusion of BOOST header file)
- fixed configure script to better differentiate between a working BOOST
  environment and a possibly problematic one
- improved ACE output for tags containing comments


2.9.45x1
--------
- fixed bug where a mapping assembly of 454 or Sanger sequences led to
  segmentation faults in some cases (*sigh*).
- fixed bug in output of assembly as HTML or TEXT format where only 21 bases
  of each contig were given (some test code had not been removed)
- fixed bug where MIRA sometimes cut back some backbone rails it thought to be
  possible chimeras. This could happen for organsism that are further apart
  from each other than just a few SNPs here and there. While this had no
  effect whatsoever on the assembly, it still was something of an unclean
  thing.
- test for consed compatibility: added newlines in ACE files to read and
  consensus tags


2.9.45
------
- to accomodate the Solexa paired-end naming scheme, CAF files now allow the
  "/" character in identifiers (like read names).
- SK:rt has been renamed to -SK:nrr and the meaning has changed (please read
  changed documentation). This gives an easier control in handling of
  repetitive sequences.
- skimming for nasty sequences (-SK:mnr) now uses the same algorithms as
  -CL:pec which are faster and better than the old ones.
- new parameter -CL:pecbph
- SKIM3 now removes some massive temporary files from the log directory
- MRMr tags renamed MNRr
- updated support files GTAGDB and consedtaglib.txt


2.9.44x7
--------
- speed up of SKIM hit reduction. Important for large eukaryotic assemblies or
  de-novo prokaryotic Solexa assemblies, reducing the time of that step from
  several hours to under one hour or even minutes.


2.9.44x6
--------
- added "solexa" as naming scheme to -GE:rns (using "/1" and "/2" to
  distinguish forward and reverse reads
- added -GE:crhf to color reads by hash frequency. Very handy for
  finishing. Needs tags "HAF0" to "HAF7" to be defined for gap4 (or consed or
  other finishing tools)
- new log file: "miralog.usedids" which logs all reads (after clipping etc.)
  which go into contig assembly
- statistics regarding the read pool are now printed out after all operations
  that might change read lengths (read extension or clipping)


2.9.44x5
--------
- added unpadded read position to "*_info_readtaglist.txt"
- -SK:pr can now be set individually by sequencing technology


2.9.44x4
--------
- bugfix in chimera search: some chimeras were not recognised, this has been
  fixed. Downside: a few more reads that are not really chimeras or were the
  info is inconclusive are now categorised as such. Should however have no
  influence on the assembly itself.


2.9.44x3
--------
- change of parameter: the "--noclipping" now takes optional technologies to
  which it should apply. E.g.: "--noclipping=454,solexa". "--noclipping" is
  equivalent to switching off all technologies.
- speed optimisation in pathfinder for de-novo assemblies with Solexa and
  SOLiD.
- bugfix: fixed some pathfinder logic where sequencing errors in repetitive
  areas led MIRA to perform alignment of reads it shouldn't have.
- bugfix: setting -SK:mchr to values >4095 led to am integer overflow and
  subsequent poor assemblies ... or no assembly at all
- bugfix: when using Solexa CER mappings on multiple backbone sequences, the
  numbering scheme led to illegal CAF files (and hence illegal gap4 databases)
- bugfix: division by zero error in statistics calculation of empty read pools


2.9.44x2
--------
- speed optimisations in new assembly engine. 5x-10x speed improvement for
  large contigs compared to 2.9.44x1
- increased threshold for megahub detection (not sure whether it's a good
  decision, must test)
- for 454 assemblies, adapted -AS:nop down to 4 for normal and 5 for accurate
  mode (improved repeat resolver and taking FLX reads as quality standard
  allows for this)


2.9.44x1
--------
- major change of assembly engine, geared towards "100% certain" contigs
  without misassembly. May lead to shorter (albeit better) contigs when no
  paired end reads are used, but leads to longer contigs for paired-ends.
  Currently very slow for large contigs (>150k reads)
- more lenient treatment of megahubs in SKIM. If possible, only skims with
  non-repetitive parts of other reads are taken.
- added searches dedicated to hunt chimeras. This was necessary as new
  assembly engine is more prone to falling into chimera traps than the old
  one.
- further improved HTML output of SNP surrounding
- SKIM now honours the -AL:mo values (minimum length of overlaps) and rejects
  overlaps below these values (important for de-novo of short reads)
- routine loading Sanger type data now gives a clearer error message if the
  file type given in -LR:snft is unknown


2.9.43
------
- fixed bug in new -CL:pec routines that led to a core dump (struck only in
  cases where not a single overlap appeared in the whole project)
- clarified docs regarding usage of ssaha(1)
- fixed problem leading to long run times and high memory requirements when
  masking of nasty repeats (-SK:mnr) was used on high coverage genomes (100x)


2.9.42x1
--------
- improved HTML output for resequenced genomes
- slightly improved logging of values when loading FASTA data
- added /proc/meminfo as dump in memory self assessments


2.9.42
------
- renamed "miraEST" to "miraSearchESTSNPs"
- internal changes to get the miraSearchESTSNPs pipeline working again in the
  2.9.x line (alpha test)
- bugfix: loading FASTA projects containing more quality entries than sequences
  led to core dump
- rewrote -CL:pec routines. Faster, and fixes errors of old version.
- change: contigs with only Solexa reads do not trigger editing of contig
  (temporary trial)
- first tests of new statistics module
- memory needs increased by 24 byte per read (12 byte on 32 bit systems) and
  one byte per raw read base
- changed default setting of poly-AT length from 10 to 12
- internal version only. Migration to gcc 4.3.2 partly done.


2.9.40
------
- new parameter: -SK:mchr to cap maximum memory in hit reduction
  algorithm. This is experimental and will need some refining.
- when no clean ends are found by proposed cutbacks, the reads are completely
  removed from the assembly. This eliminates short reads (i.e. Solexa) with
  too many errors and which aren't really useful anyway.
- Using -AL:shme led to a parsing error. Fixed.
- Solexa reads now do not need anymore a minimum left clip to be set, this is
  handled internally
- miramem now gives a better estimate for mapping of Solexa reads (the old
  values were way too high)
- miramem no tries to split memory needs into "unavoidable" (for sequencing
  data etc.) and "tunable" (via a number of parameters)
- -SB:bbq now defaults to '30'
- most error messages now dumped to STDOUT instead of STDERR


2.9.39
------
- in mapping assemblies, repetitive reads are now distributed evenly and not
  stochastically distributed over the backbone repeats
- mapping assemblies with Solexa now have some adjusted default parameters for
  "normal" and "accurate" levels. They run a bit slower but will squeeze a
  maximum out of your data.
- read clustering now temporarily needs more memory, but runs in a few seconds
  instead of hours for projects with 10 million reads
- new parameter -AL:shme (a temporary hack to handle Solexa reads more
  thoroughly)
- to counter a current defficiency of the Solexa technology, a new clipping
  filter for Solexa data now filters out reads that have stretches of 20 or
  more "A" bases or stretches of 12 or more "A" bases and more than 80% "A" in
  total.
- on "out-of-memory" errors, MIRA now dumps a self assessment on where the
  memory went to get an idea what really happened. Note 1: this is bound to
  happen only with eukaryots or on very small machines. Note 2: development
  versions of MIRA by default dump some assements also during the assembly.
- documented -OUT:sssip:stsip (which appeared in 2.9.12, my apologies)
- changed documentation for 454 assembly to point at publicly available data
  instead of the spneu project (which put too much strain on my website).
- renamed -CL:prc to -CL:pec to reflect it's use on both ends of a read


2.9.38x1
--------
- -CL:prc now also clips left (will have to rename that option). This catches
   very efficiently vector leftovers in Sanger reads and adaptor leftovers in
   454 reads (which also can occur there).
- -CL:prc now also clips when a non-ACGT base is at the ends
- bugfix: saving as gap4 directory did not save the first contig due to wrong
  handling of directory creation.
- bugfix: convert_project now sets the minimum coverage to 1 to circumvent a
  quirk in the computation of "Large contigs" of the assembly info
  display. Better fix in the future.
- version 1a: testing new pathfinder algorithm enabled


2.9.37
------
- new option: -CL:prc (propose right clip). This is a new strategy to ensure a
  good "high confidence region" (hcr) in reads, basically eliminating all junk
  at the 3' end of reads. Extremely effective, but should not be used for very
  low coverage data or for EST projects.
  This option is now default for genome assemblies in "normal" or "accurate"
  mode.


2.9.36
------
- renamed -AS:urdufrd:urdrdct to -AS:ard:ardct
- added -AS:ardml:ardgl. This allows for a better control of which reads are
  defined as repeats.
- added -AS:klrs. Needs testing is not switched on by default.
- bugfix: number of large contigs was reported too high in the report of the
  assembly ... because of a really dumb bug in the statistics calculation
  routine. This had no effect on the assembly itself, just on the
  *_info_asembly.txt report and also on the summary given after the usage of
  "convert_project".
- bugfix: SSAHA clips were wrongly logged to file
- change: log file with clips more verbose
- change: 454 reads without explicit forward/reverse naming scheme
  (e.g. "somename" instead of "somename.f") are now considered to be forward


2.9.35x2
--------
- when running the SKIM in parallel threads, MIRA can give 
  different results when started with the same data and same
  arguments. The effect is now reduced (it is still present), but at the price
  of a table loaded after SKIM ran through now being 25% larger, but this can
  not be helped.
- a few fixes in "convert_project" to allow conversion of assemblies in CAF
  format into clippedfasta and maskedfasta (was previously allowed only for
  single reads)
- typo fix: -OUT:rrol:rld were shown as sequencing type dependent while they
  are not.


2.9.35x1
--------
- CAF files with 454 data now contain the necessary info to allow gap4 opening
  the flowgrams. Works only for reads that are NOT paired-end.
- slight tweak in the pathfinder that should enhance the assembly with
  paired-end in a few cases
- changed sff_extract so that it runs again with the Python 2.4 series


2.9.34
------
- bugfix: fixed bug while reading quoted text for "Clone", "Staden_id" and
  "Template" lines in CAF. This affected mostly users of convert_project and
  users using CAF as input format for mira.


2.9.33
------
- new output file: wiggle (-OUT:orw). Can be loaded into the IGB viewer
  together with a GFF or FASTA of the backbone sequence(s). Very useful for
  resequencing experiments / mapping assemblies with 454 / Solexa data, when
  the "view coverage" of gap4 gets really really slow.
- convert_project can now convert to multiple targets at once (multiple -t)
- fixed wrong reporting of 454 reads without clips in the statistics display
  after loading of reads.
- improved sff_extract to handle paired-end reads, rewrote 454 manual to
  reflect this.


2.9.32
------
- tweak: --job=...,est,... now switches off -CL:emrc per default


2.9.31
------
- reworked info file for contig statistics: contig lengths are reported for
  the ungapped contigs; added GC content (removed A/C/G/T counts); format is
  now more easily readable but can still be easily parsed.
- fixed bug where reads with very short "good" parts could loose ther right
  vector clip when using -CL:emrc
- bugfix: progress indicator for files >2GB did not work correctly


2.9.30
------
- added warn message if skim was parametrised in a way that make it run slow
- optimised some input/output of temporary files to be faster (using C instead
  of C++ functions *sigh*)
- name scheme of reads now allows for ":" in names (to accept the original
  Solexa name scheme)
- tweak: a mapping assembly that has Solexa will now generate filter more
  strongly during the SKIm pahse, saving some memory afterwards in the whole
  assembly
- bugfix: -AS:bdq did work only with partial FASTA quality files, not with
  empty or non-existing files
- bugfix: marking repeats of Solexa reads did not completely honour -CO:mrpg,
  fixed
- bugfix: when performing Solexa mappings, MIRA sometimes created almost empty
  CER (coverage equivalent reads) without strain information.


2.9.29x6
--------
- tweak: adjusted parameters for Solexa mapping to be more lenient in
  alignments
- bugfix: --noclipping now also correctly switches of -CL:emrc
- bugfix: --job in parameter files was still not correctly parsed in some
  cases (bug in flex *sigh*)
- bugfix: threaded skim sometimes did not exit when given less sequences
  than threads*5000


2.9.29x5
--------
- removed the "sffinfo2mirafiles.tcl" script from the
  distribution. "sff_extract" from Jose Blanca (in the 3rd party package) is
  taking over this part: more versatile, faster and removes the need of for
  the sff* tools from Roche.
- MIRA now stops if the ratio of megahubs is larger than -SK:mmhr
- tweak: EST mode now does not enforce minimum right clip per default
- new tag: MCVc (Missing CoVerage in Consensus). Set when a strain has no
  coverage (previously UNSc and UNSr were set, now they are set only when
  'unsure').
- bugfix: non-paired-end 454 data read from CAF was not recognised as
  non-paired-end during -CO:emrc, fixed.
- bugfix: --job in parameter files was not correctly parsed
- bugfix: when -OUT:sssip was used, parts of the singlets were still assigned
  to the debris file.


2.9.29x4
--------
- the cutback strategy for 454 reads introduced in 2929x1 has been eased a
  bit: paired-end reads do not get cut back and if read lengths would fall
  below the minumum required length, they don't get cut back neither.
- bugfix: the bad sequence search was not performed if the minimum left clip
  was set to "no"
- MACHINE_TYPE, PROGRAM_ID and STRAIN are now read from TRACEINFO XML files.
- fixed a couple of bugs that led to an abort of MIRA when writing SNP files.
  Bug struck only on very rare boundary cases.


2.9.29x3
--------
- implemented a couple of memory reduction strategies for the read
  objects. This reduces overhead of every read by almost 30% (264 instead of
  368) and additionally saves memory of cached values (strain, basecaller,
  machine type, paths etc.). This should also reduce memory fragmentation a
  bit.
  In a typical 454 project with 1 million reads, this amounts to 208-400 MB
  savings of RAM.
- miramem now knows about 454 Titanium reads
- convert_project has new command line parameter: -r


2.9.29x2
--------
- additional algorithms to search and mark repeats marker bases that existing
  routines missed in 454 data.


2.9.29x1
--------
- MIRA now uses full overlap graph repeat resolving algorithms which leads to
  better and quicker resolving of repeats in bacteria. May be slower for
  eukaryotes, more tests needed.
- new clipping options: -CL:emrc:mrcr:smrc
- for 454 reads, MIRA now follows a strategy of cut back first (-CL:emrc),
  uncover afterwards via read extension. Highly recommended.
- default parameter -CO:mrpg=5 for repeat marker base detection in 454 data was
  to lax, changed back to 4.
- fixed bug: when mapping microread data (Solexa, SOLiD), -SB:sbuip was
  wrongly interpreted and de-novo algorithm started instead of mapping
  (error introduced in 2.9.28x4)
- change: when not being able to delete a temporary log file, MIRA now gives
  a warning but does not abort


2.9.28x7
--------
- added quality information of consensus sequence to output of CAF files.


2.9.28x6
--------
- Premiere for MIRA: multi-threading makes its appearance. At the moment only
  for the SKIM algorithm as it's the easiest part and no adverse effects
  are expected.
  New parameter -SK:not is for controlling the number of threads.
- Test: MIRA now saves more information on failed alignments to build a better
  overlap graph in following passes. The overall assembly quality gains, but
  memory consumption rises unpredictably. This may become a problem for highly
  repetitive genomes of eukaryotic size. To be monitored.
- the rawhashhit log file is not written anymore as it was useful only for
  debugging and just ate memory and time of SKIM.
- bugfix: the new read mapping chooser sometimes led to an abort() of the
  process (error introduced in 2.9.28x4)


2.9.28x5
--------
- renamed 'est_splitsplices' of the -AL:egpl parameter to 'reject_codongaps'
- when 454 data is used via the --job=...,454,... switch,
  -AL:egp=yes:egpl=reject_codongaps are now set for *all* technologies


2.9.28x4
--------
- first version which allows Solexa de-novo. Albeit *very* slow at the
  moment, do not use for anything else than bacteria (1 week of computation or
  more, sorry).
- new functionality: MIRA now marks IUPAC positions in the consensus as tag
  "IUPc"
- the info_assembly.txt file now gives info on the number of positions in the
  consensus where sequencing methods disagree.
- bugfix in clipping: when reads have no ancillary data (and this no good left
  clip), the clipping of bad quality stretches could lead to the complete read
  being clipped if the bad quality on the left was long enough
- bugfix: for 454 data, -CO:mrpg and -CO:mgqrt were not honoured but some
  fixed values used instead.
- bugfix: average total coverage of contigs was wrongly reported in statistics
  and furthermore only as integer, not as double
- bugfix: -OUT:sssip:stsip could not be set for sequencing technologies other
  than Sanger.
- optimisation: when mapping against multiple backbones, reads will now be
  mapped to the best matching backbone instead of suboptimal mapping to
  backbones earlier in the list.


2.9.28x3
--------
!!!! Do not use this version with Solexa data, some code changes are not
  completed !!!!
- changed graph pruning algorithm to work less aggressively so that
  454/Sanger hybrid situations, where there's a low Sanger coverage (0.5x to
  2x), now should work better than before
- fixed bug in assembly info which led to wrong information being displayed
  which struck when using 454 data
- reduced memory requirements of one of the main overlap storage tables by
  10%


2.9.28x2
--------
- fixed ugly bug in new assembly info routines that led to an abort of the
  assembly.


2.9.28x1
--------
- new file in *_info directory: <projectname>_info_assembly.txt. This file
  gives basic information on how the assembly went (assembled reads, contig
  sizes, N50/N90/N95, coverage, qualities, possible problems)
- convert_project now dumps the same info as above on CAF input
- for EST assemblies: -job=est now uses per default a poly-AT clipping that
  preserves the poly-A/T signal
- for EST assemblies: renamed -GE:ess to -GE:esps
- for EST assemblies: added -job=esps[1-3]
- made tagsnp working again (for testing)


2.9.27x2
--------
- new parameter to force consensus to be A, C, G or T and not a IUPAC
  code (-CO:fnicpst) 
- enhanced handling of partly masked reads.
- implemented alternative SKIM repeat threshold calculation to deal with
  highly repetitive eukaryotes (temp. fix, needs revisiting later)
- temporarily took out default -SK:mnr again


2.9.27x1
--------
- fixed bug in ACE output which was introduced in 26x6
- fixed small bug when "--job=mapping" was used (introduced in 2924x3)
- activated SKIM routines to mask nasty repeats, added -SK:mnr:rt
- massively reduced disk usage as MIRA can now remove unused log files and the
  complete log directory on request (added -OUT:rrol:rld)
- started documentation for log files written by MIRA


2.9.26x6
--------
- fixed bug in output of read and consensus tag comments in ACE files
- fixed bug in output of BS lines of ACE files
- added error messages naming the faultive read during loading of CAF files
- added -AS:urdcm:urdufrd:urdrdct
- while building contigs using backbones, MIRA now tries to guess memory usage
  and uses preallocation to reduce memory footprint. Though the guesstimate
  may be wrong a few times which then leads to increased memory usage.
- added tag type MRMr (MIRA Repeat Marker)
- SKIM now honours "MRMr" and "FPaS" tags in reads and does not use these
  stretches to find potential overlaps.
- SKIM now adapts dynamically hashes it needs to save to non-ACGT bases
  occuring in sequences. This leads to slightly improved detection of possible
  overlaps in sequences with "N" or other IUPAC codes


2.9.26x5
--------
- fixed bug that was introduced in 26x2 where sometimes during a mapping
  assembly, all contig positions were tagged as unsure.
- fixed bug that -FN:xtii could not be set for sequencing types other than
  Sanger.
- fixed bug: quality clip and clip of masked bases were performed for all
  sequencing technologies even when only one requested it.


2.9.26x4
--------
- added log file for clippings on load
- fixed bug for minimum left clip function
- fixed small bug in "make install"
- testing new workaround for linking on MacOS X (10.5) *sigh*


2.9.26x3
--------
- fixed nasty misconfiguration: --job=accurate,454 did not switch on 454
  editing
- fixed nasty misconfiguration: --job=est,454 did not use optimal
  parametrization for 454 EST data


2.9.26x2
--------
- major change for mapping assemblies: strains that do not cover the backbones
  will get the '@' as consensus character. This also appears in result files
  (FASTAs) specific to the different strains! Therefore, some post-processing
  may now be needed.
- revamped and improved strain difference analysis routines (for mapping
  assemblies)


2.9.26x1
--------
- tweaked consensus routines for hybrid assemblies
- added the St. Louis read naming convention as read naming scheme
  (-LR:rns=stlouis)
- updated and expanded some docs (mira manual, usage and 454)
- paired-end reads that have a partner now do not get thrown out at the
  beginning of the assembly on the minumum length criterium (-AS:mrl). This is
  to accomodate 454 paired-ends where one read of the pair sometimes might be
  really, really short. Same thing applies for Solexa.


2.9.25
------
- fixed small bug for mapping assembly: reads smaller -PF:bqoml were not
  mapped at all.
- the automatic error editing routines for overcalls for new sequencing
  technologies data previously worked only with 454 data. They now edit in the
  following combinations:
    1) 454 only
    2) Solexa only
    3) Hybrid assembly of Sanger with (454 and / or Solexa)
       or 454 with Solexa
- the contig statistics file now contains a column for non-covered backbone
  positions (important for backbone assemblies)
- fixed small bug: sometimes singlets were saved in projects even when not
  requested (-OUT:sssip:stsip)
- fixed anoying bug: MIRA (and "convert_project" lost information about
  backbones or merged short reads
- -OUT:sssip:stsip can now be set dependend on sequencing technology
- fixed bug: Solexa consensus sometimes chose the wrong base in cases of
  conflict


2.9.24x3
--------
- added switches and documentation for uniform read distribution
  (-AS:urd:urdsip).
- improved uniform read distribution a bit and made it default for genome
  assemblies (when there is no Solexa data, not tested on that yet)
- added "miramem" as program call to help estimate memory needed for an
  assembly


2.9.24x2
--------
- test version with trial for uniform read distribution
- fixed bug that led to MIRA aborting contig assembly in rare cases (triggered
  through quite repetitive sequences). Bug was introduced after 2.9.17.
- take back too daring optimisation for Solexa mappings (did sometimes not map
  important data)
- merged short reads now get strain information attached (only one strain at
  the moment)
- added -m to convert_project


2.9.24x1
--------
Bugfixes/tweaks when using more than one strain (bringing back and improving
functionality that was lost in the 2.8.x -> 2.9.x changes)
- added -SB:bsnffa:brfs
- added --notraceinfo quickswitch
- bugfix: searching for SNPs now only done when having multiple strains in
  assembly
- testing: now also setting tags for "weak" SNPs (for catching indels with 454
  reads)
- bugfix: now correctly aligns Solexa reads on backbones containing gaps (as
  encountered in CAFs when using a given strain as rail constructor).
- a few more bugfixes and tweaks concerning assembly with multiple strains
- tweak: gap2caf sets "clone" information to 'unknown'. MIRA now treat this as
  "not set" instead of having an additional "unknown" strain.


2.9.23
------
- fixed bug that led to quality clipping even if -CL:qc was no in cases the
  clip to masked characters (-CL:mc) was active
- fixed bug that made convert_project "forget" to write strain info back to
  caf


2.9.22x4
--------
- improved again tagging of SNPs when using multiple strains and sequencing
  technologies (now also adds "medium" SNPs)
- fixed bug that sometimes led to additional rounds of repeat disentangling
  alignments not being called


2.9.22x3
--------
- tweaked/improved Solexa base calling
- improved tagging of SNPs when using multiple strains and sequencing
  technologies
- added "-noclipping" quickswitch that switches of every clipping option
- added "-lowqualitydata" quickswitch
- added "-notraceinfo" quickswitch
- added -CO:mroir
- changed tagging/clipping of poly-A signal to perform as full blown clipping
  routine. Renamed -DP:tpae and related options to -CL:cpat and related
- bugfix / fallout from changes in 2.9.20x1: the clipping routines for the
  following options now honour the sequencing technology specific settings:
  -CL:msvs:emlc:bsqc:qc:mbc:cpat
- bugfix: -OUT:oet* is now working again
- bugfix: when using multiple strains, the new consensus routine sometimes
  returned '?'.
- re-activated "miraclip" (needs testing)
- re-activated "mirapre" (needs real testing)
- documented -CL:bsqc
- brought most of the main documentation up-to-date to 2.9.22


2.9.22x2
--------
- tweaked routines for calculation of Solexa consensus
- bugfix: statistics calculation of Solexa data was not correct in 2.9.22x1
- small fixes around the code


2.9.22x1
--------
- added support for simple forward / reverse read naming scheme
- fixed bug in template strand assignment for Sanger read naming scheme
- revamped contig statistics in logfile (info file will follow in future)
- renamed "RT=" to "ST=" in MINF read tags
- changed sequencing technology "454GS20" to "454GS" in MINF tags


2.9.21
------
- speed improvement of the SKIM algorithm when only mapping reads to (a)
  backbone(s). Reduced complexity from quadratic to linear, SKIMing a few
  millions Solexa reads against a backbone now just takes a few minutes
  instead of one hour or more.
- speed improvement of mapping phase, mapping a few million Solexa reads now
  takes a few minutes per round instead of hours.
- automatic Solexa read clip back mechanism activated that honours the quality
  of mismatches as well as the necessity of having all data at the given
  mapping place.
- added a small hack to be able to use the full Solexa read length and still
  hide the MINF tags in GAP4: reads now get a N added as first base
  just to get clipped away in quality clipping (MINF tags should be replaced
  by notes when I have time to do that)
- the CAF loader now gracefully handles sequences without quality values
  (although this should not happen) by setting default qualities
- tweaked standard parameters for the different read types
- small fix to make .ace files immediately readable by consed
- a number of smaller bugfixes (like correcting typos etc.pp)


2.9.20x2
--------
- bugfix of a rare error during alignment of reads to contig (new from
  2.9.19x1)
- can now load Solexa quality scores in FASTA quality files and convert them
  to phred-style quality values (new parameter -LR:ssiqf)
- statistics of reads in assembly are now given for each sequencing type
  separately
- added COMMON_SETTINGS as alias to SANGER_SETTINGS, should help to clarify
  things a bit in parameter files
- removed -horrid
- replaced STRM tags with STMS and STMU
- added possibility to set different input / output project names
  (-projectin= and -projectout=), the (still existing) -project= is simply a
  combination of in/out
- improved repeat tagging when different seqencing technologies are involved.


2.9.20x1 (do not use with Solexa data)
--------------------------------------
- major rework of parameters information display, no separate info for each
  sequencing type is shown where appropriate
- added SANGER_SETTINGS, 454_..., SOLEXA_..., SOLID_... This allows to set
  all parameters for all sequencing types in one file (or the command line).
  Or distribute the settings across different files, whatever one wants.
- removed hack to load parameters specifically for 454 data (file
  "454params.par") as functionality is given by the ..._SETTINGS above.
- new -LR category (merged -454 and -SR, together with some -GE)


2.9.19x1 (do not use with Solexa data)
--------------------------------------
- fixed minor bug that led to bases sometimes being aligned more against
  gap columns than against consensus bases
- minor changes in quality computation towards end of reads
- changed quality computation for consensus with multiple sequencing types:
  now not best quality only, but additive quality
- fixed bug when computing Solexa only consensus (appeared in 2.9.18)


2.9.18
------
Preparing for simple Solexa / SOLiD mapping assembly:
- added -SR
- added -PF:bqoml
- added -SB:bro

- fixed bug that led to inclusion of 100% potential overlaps smaller than
  minimum allowed overlap (had probably only very small effect on assemblies)
  Bug appeared in 2.9.17 (SW not being recomputed).


2.9.17
------
- improved template handling: TRACE_END is now read from TRACEINFO XML
- improved template handling: "Strand" is now read/written from/to CAFs,
  leading to correct information in Staden projects when caf2gap is used
- increased speed of SW alignment phase: perfect matches are not computed
  again. Saves 50-70% of SW alignments for a typical project with 454
  GS20 data, with Sanger data it's still ~25%.
- increased speed of adding aligned reads to contigs. Important for contigs
  >50k reads, saving ~20-30% time.
- fixed bug in routine that should have sped up endgame of an assembly (came
  in in 2.9.14b) but led to inconsistent exclusion of reads.
- added hack to load parameters specifically for 454 data (file
  "454params.par") (later removed in 2.9.20x1)
- (internal: added timing measurements for contig and pathfinder objects to
  search for unfriendly runtime behaviour)


2.9.16x1
--------
Experimental version for improved hybrid assemblies
- fixed bug that led to short contigs when performing a hybrid assembly with
  low Sanger coverage
- added -PF:uqr:qrml1:qrms1:qrml2:qrms2


2.9.15
------
- fixed bug in SequenceVector handling of CAF reading routine that was
  triggered by 454 type data


2.9.14b
-------
- rewrote large parts of the 454 assembly tutorial
- reworked the spneut4demo_assemblies package
- added "sffinfo2mirafiles.tcl" script to the distribution
- adapted a few internal parameters for 454 assembly (a bit more stringent,
  uses a bit less memory and is slightly faster), using similar standard
  parameters as Newbler.
- routine to speed up of endgame of an assembly: intermediate singlets are
  converted into debris. Speeds up assembly of SpneumoniaeT4 with a 7-3
  scheme from ~47hrs to ~21hrs runtime.
- de-activated old reads-only editing routines (-454:soer:soemq)
- renamed -FILE (-FI) to -FILENAME (-FN) to reflect internal structure
- fixed output file name of miraclip to name given in parameters (or
  constructed from project name)


2.9.14a
-------
Feature freeze for 2.9.15 (target: mira easily usable for 454 and 454 /
Sanger)
- moved -CO:dismin:dismax to -AS:tismin:tismax
- change in behaviour: template insert sizes now do not get assigned anymore
  by default, only on request (-AS:tismin:tismax being unequal to -1). That
  is, reads that do not have this information as ancillary data will not get
  default insert sizes assigned unless expressedly wished so.
  This fixes the bug of 454 reads without template information getting insert
  sizes assigned.


2.9.14
------
- new parameters -CLIP:bsqc*
- fixed bug: multicopy reads were not detected after early SKIM phase (present
  since 2.9.12)
- small adaptions for SKIM hit reduction for 454 only assemblies


2.9.13
------
major adaptation and feature enhancement of convert_project to changes after
2.9.8 (not finished, but useable again)

- bugfix: convert_project and other tools now keep name of contig and do not
  rename this to stdname_... anymore
- bugfix: convert_project and other tools now do not recalculate the consensus
  when loading from CAF. Note however that consensus qualities must be
  recalculated when they are not stored in CAF files. This recalculation may
  lead to slightly different quality values. 
- convert_project and other tools now keep order of reads when writing back to
  CAF files
- when loading a CAF with contigs, convert_project now only needs enough
  memory to convert one contig at a time, not the complete project
- workaround: caf2gap failed to convert CAF files where the contig name is
  equal to the name of a read


2.9.12
------
- bugfix: insert size standard deviation was not read due type from XML
  TRACEINFO files (and hence the standard of 500 used, which was sometimes not
  enough for libraries >3kb)
- reduced memory footprint of pathfinder algorithm, important for high
  coverage 454 areas.
- reduced memory footprint of alignment storage by 16% on 64 bit architectures
  (less for 32 bit architectures)
- both reductions now allow de-novo hybrid assembly of S.pneumoniae (1.1
  million reads) with 4 GB RAM (and some 2GB free swap).
- speed increase by factor 2-5 (depending on repetitive area) in pathfinder
  algorithm for 454 reads. Effectively almost halving total time needed in
  hybrid and highly stacked 454 assemblies
- improved pathfinder algorithm for better bridging or repeats
- storage of singlets in project results can now be controlled by
  -OUT:sssip:stsip


2.9.11
------
- bugfix: cut down overzealous editing of 454 reads in misassembled parts of a
  contig
- enhanced internal handling of sequencing data types. Routines now use
  dedicated parameter sets for each type (cannot be set from command line or
  parameter file yet)


2.9.10
------
- major fixes in alignment of contigs which allows better hybrid
  assemblies (more reads added)
- major rework of 454 read editing. Kicks out most of the obvious sequencing
  errors and now pre-assembly editing reads-only is not required anymore.
- major new memory saving option: -GE:kcim (not compatible with spoiler
  detection -AS:sd)
- new routines to thin out overlap graph which reduces number of initial
  Smith-Waterman overlap alignments by 80%-90% (increasing speed in this part
  by 5x - 10x). Drawback (mainly 454 data): really highly repetitive areas
  with complicated solution space will not get optimally solved and more than
  expected reads of these areas will turn into singlets.


2.9.9
-----
- MIRA now support merging results from a SSAHA vector screen run. This makes
  you basically independent from any other commercial or license-requiring
  vector screening software. For Sanger reads, a combination of "lucy"
  and "ssaha" together with this parameter should do the trick. For
  reads coming from 454 pyro-sequencing, \Cmd{ssaha}{1} and this parameter
  will work very well. New parameters:
  -CL:msvs:msvsgs:msvsmfg:msvsmeg:msvssfc:msvssec and -FI:svsi
- along with the above: a new mira program "miraCLIP" has been created that
  just clips data from loaded files and dumps them to CAF format. The
  "miraPRE" program should now be used only for a first repeat-disentangling
  assembly.
- renamed -CL:pvc to -CL:pvlc to make function clearer
- Work in progress: new functionality that reduces number of hits that must be
  checked by SW alignment when working with 454 data, especially useful for
  hybrid Sanger / 454 assemblies. First tests very promising
- the "|" characters in the name of reads in fasta files are now kept and not
  replaced anymore with "_", only in the EXP file names written to disk they
  are replaced
- optimised memory handling when loading CAF files, leading to decreased
  memory footprint of sequences loaded via CAF (important for millions of
  reads)


2.9.8b,c,d,e,f
--------------
- polishing of the MIRA build process, added flex version check, added
  --enable-static flag to ./configure
- fixed build process for MacOS X (Darwin)
- bugfix in alignment routine: hitting band encasing was not discovered in
  some cases (present since adding feature in 2.9.2)
- fixed minor code ambiguouities
- bugfix: minimum left clips (-CL:emlc) were not performed for 454 data
- added mira_454dev help file as a preliminary guide for 454 assembly 


2.9.8
-----
- new program: "miraPRE". miraPRE is a preprocessing step that allows to
  perform the preprocessing of reads (clipping, read extension, simple 454
  editing) together with a first "reconnaissance" assembly of the most
  repetitive regions. The following "real" assembly is then faster (allowing
  for playing around with some more options) and more accurate for repeats
- bugfix for hybrid assemblies: repeats found were not correctly accounted
  for, leading to the assembler not correctly recognising when to re-assemble
- minor bugfix: the logfiles for rejected alignments contained binary data


2.9.7
-----
- bugfix: 454 reads now are not put through the possible sequence vector
  clipping routines
- bugfix: assemblies with several backbone sequences could lead to the
  assembler aborting with an error. Fixed. Present since 2.7.5, but different
  bug than the one fixed in 2.7.6.
- improved new consensus computation routines introduced in the 2.9 series
  (needed because of hybrid assemblies) to better handle aberrant Sanger cases
- improved consensus computation for hybrid Sanger / 454 assemblies
- improved loading speed of qualities in fasta files where the reads are not
  in the same order as in the sequence files (mostly noticed with 454 data
  files having millions of reads)
- renamed output directories to <projectname>_d_info, <projectname>_d_results
  and <projectname>_d_log
- small change in output behaviour: contigs are now put first in output files
  (CAF, GAP4DA, FASTA etc.pp), then follow singlets
- change in behaviour: debris are now left out from result files of an
  assembly. Debris are reads that are too short or do not align to any other
  read in the data set. Also, 454 reads that could not be assembled into a
  contig are treated as debris (even if they potentially aligned to other
  reads in the assembly).
- automatically switching off output of GAP4DA format if 454 type data is
  present (you *really* do not want millions of files in a directory)


2.9.6
-----
- adapted internal multicopy detection for hybrid assemblies
- adapted simple 454 overcall editing for hybrid assemblies
- enhanced XML traceinfo support for files directly from the NCBI trace
  archive:
   - support for XML "ti": now, XMLs & FASTAs from NCBI need not to be
     rewritten (changing names etc.)
   - support for XML trace_type_code (now possible to mix Sanger type & 454
     type in one fasta and XML traceinfo respectively)
- fixed bug in XML traceinfo routines: XML elements with uppercase letters
  were not recognised
- fixed ugly bug that led in very rare cases to suboptimal or missed
  alignments in banded SW. Bug triggered by short read data (454 type),
  present since MIRA V1.1.1, correction attempt in V2.2.2/2.3.3 only
  partly successful (ooops).
- fixed bug that led to slow SW band alignments in development version
  (introduced through bugfix in V2.9.2)


2.9.5
-----
- tweaked internal memory handling (STL), reducing memory footprint of MIRA


2.9.4
-----
- fixed bug in tag type set for SNPs between strains (introduced in 2.9.3)
- improved repeat resolving when using low number of passes (e.g. only 2) or
  for highly complicated repetitive projects by adding selective SW alignment
  iteration after each pass.
- improved interaction between contig building and 454 contig editing process,
  leading to less building loops needed to achieve the same result


2.9.3
-----
- introduced "Carbon-copy Repeat Marker in Reads" (CRMr) which tremendously
  help assembling repetitive 454 sequences (also good for Sanger sequences)
- reactivated Sanger repeat recognition with new repeat handling routines
- fixed bug: when loading sequences with gaps (like in assembled CAF
  projects), existing gaps in reads were not removed prior to re-assembly
- fixed bug that prevented mapping assemblies against backbones to be fast


2.9.2
-----
- added -454:soer:soemq for "simple" editing of errors in 454 reads.
- MIRA now stores additional information needed for assemblies of strains
  and/or 454 data in CAF and EXP files in a way that resist to different
  transformations to and from gap4 databases (MINF tags)
- bugfix when loading CAF: some attributes were not reset, leading to some
  reads having attributes of preceding reads in CAF file.
- fixed bugs that sometimes led to suboptimal alignments when building contig
  alignments. This became apparent with assembly of highly stacked 454 data
- first working version of repeat discovery in 454 data
- first working version of "tricky overcall editing" in 454 data (these things
  are responsible for most of the frameshift errors)


2.9.1
-----
- test activation of 454 read-only-editing routines



Change file for MIRA 2.8.3
==========================


2.8.3
-----
- fixed bug in XML traceinfo routines: XML elements with uppercase letters
  were not recognised (backport from the 2.9 development line)


2.8.2
-----
- polishing of the MIRA build process, added flex version check, added
  --enable-static flag to ./configure


2.8.1
-----
- fixed bug that prevented mapping assemblies against backbones to be fast
  (backported from 2.9.3)
- bugfix when loading CAF: some attributes were not reset, leading some reads
  having attributes of preceding reads in CAF file (backported from 2.9.2)


2.8.0
-----
- replace tcl dependency in compile process with perl dependency


2.6.x -> 2.7.x -> 2.8.0rc2
--------------------------

The 2.8.x series of MIRA is an intermediate step towards MIRA 3.0. However,
two major highlights (beside the usual stream of small improvements and
bugfixes) justify a new production release:

1) Important speedups in a few central places (Reads, Contigs). Additionally,
   the all-against-all read comparison algorithm (SKIM) has been speeded-up by
   a factor of >60 (for large number of reads).
2) MIRA is going Open Source! More specifically, MIRA is being put under the
   GPL (version 2) as I kindly received the authorisation from both the DKFZ
   Heidelberg (Deutsches Krebsforschungszentrum, German Cancer Research
   Center) and from Thomas Pfisterer (the author of the EdIt part in MIRA) to
   release the code.

Please note that the 2.8 line does still not officially support assembly of
454 data. These routines are still under development and will be made
available in the 2.9 development line.

Important notice: a few parameter switches were changed since the 2.6.x
releases, existing parameter files may have to be changed. Please consult the
documentation for the new names of the parameters.



Changes in detail:
==================

2.7.8
-----
- Major bugfix: assembling with more than one strain without backbone produced
  less than sub-optimal solutions.


2.7.7
-----
- internal changes
- fixed error in assembly class in handling bases tagged as SIOx within a
  strain. (had only minor repercussion in assemblies)
- fixed bug in skim output: permbans were not updated correctly for the
  summary. (had no effect whatsover on assemblies, just in the printed skim
  summary of the MIRA log)


2.7.6
-----
- smaller internal changes (removed deprecated strstream constructs, removed
  old code)
- also put EdIt and other code from Thomas under GPL
- fixed bug when loading backbones sequences that have no sequence
- fixed bug when assembling with more than one backbone sequence (introduced
  in 2.7.5)
- added -AS:bdq
- brought documentation up-to-date


2.7.5
-----
- Larger internal changes. Reduced memory footprint for alignment checks:
  results are now temporarily written to disks instead of being kept in
  memory. Useful when working with millions of reads.
- speedups in contig: moved some redundant costly internal checks
  into versions compiled only for special bughunting
- removed -SK:im:mc
- added -SK:bph:hss:mhim
- new log file containing SKIM raw hash hit numbers in log directory
- new log file containing simple readpool info in log directory
- squashed a small bug in alignment code that sneaked in in 2.7.2
- switched miraEST on again (switched off in 2.7.4). SNP analysis still the
  old one though.
- the pathfinder object is now faster for backbone assembly.
- put MIRA code under GPL


2.7.4
-----
- added -454:c454cq (not public yet)
- change: MIRA now writes files into three direcorties below the starting
  directory: <projectname>_log, <projectname>_info and
  <projectname>_results. This helps to keep things a bit cleaner in the
  directory.
- due to the above: temporarily switched off miraEST (also wrt the fact that
  SNP analysis is going to undergo a major rehaul)
- speeded up endgame (coping with remaining singlets) of assemblies that have
  a large number of reads.
- *major* speedup of SKIM (all against all comparison)
  routine. E.g. SKIMming of 53,000 reads now takes a minute instead of 62
  minutes.
- reduced memory footprint for SKIM: results are now written to disks instead
  of being kept in memory. Useful when working with millions of reads.

2.7.3
-----
- first trial version that assembles 454 consensus data (not real 454 data
  yet)
- changed pathfinder strategy. Now uses a less aggressive way of determining
  next read to add. Should improve all "difficult" assemblies with many
  repeats a bit
- reworked/changed all existing -454 parameters
- Change: there is no more default source for the reads to be loaded. It now
  must be explicitly set, either via -GE:lj or the quickswitches like -fasta
  etc.
- Name change: main assembly 'loops' are now called assembly 'passes' for
  better distinction to PRMB break loops. Renamed -AS:nol:sel:sdllo to
  -AS:nop:sep:sdlpo, -SB:sbuil to -SB:sbuip, -DP:feil:leil to -DP:feip:leip
- MIRA now writes clear parameter parsing error messages (if needed) on
  startup

2.7.2
-----
- speedups in assembly & output: moved some redundant costly internal checks
  into versions compiled only for special bughunting
- changed consensus quality computation (faster). Results are mostly the same
  or a tad higher than for the old routines (a tad lower for quality values <=
  10 or so).


2.7.1
-----
- fixed small bug in handling of tags for alignments


2.7.0
-----
- initial takeover from 2.6.0


2.4.x -> 2.5.x -> 2.6.0
-----------------------

Main development focus for the 2.6 production release of the MIRA assembly
tools was to realise improvements in speed and memory footprint compared to
the 2.4 line. All changes were extensively tested in the 2.5.x development
versions and were ported to the new 2.6.0 production version.

Highlights:
- new, easy to use-and-combine parameter switches for predefined tasks: quick
  switches. (also called dwim: Do-What-I-Mean switches)
- constant memory SKIM routine for fast all against all overlap checking. This
  was needed when assembling larger bacteria or lower eukaryotes in a limited
  amount of memory. As bonus, this new SKIM is 40% to 60% faster than the old
  one.
- reduced memory footprint for stored alignments. 37% reduction in this part,
  important for big assemblies
- enhanced read extension routines. The new routines have are now quite
  efficient in extending reads as much as possible while leaving really bad
  quality parts untouched
- new type of output: GBF (GenBank file). Extremely useful when performing
  assemblies against a backbone (read mapping strategy) which in itself may be
  a GenBank file containing features.
- enhanced convert_project utility. Converts more assembly file formats
- a number of small bugfixes and other improvements

Please note that the 2.6 line does not officially support assembly of 454
data. These routines are still under development and will be made available in
the 2.7 development line.

Important notice: a few parameter switches were changed since the 2.4.x
releases, existing parameter files may have to be changed. Please consult the
documentation for the new names of the parameters.

Focus for next development cycle:
- multithreading
- 454 data


2.2.8 to 2.4.0
--------------
The new major 2.4 release line of the MIRA assembly tools opens a
whole new set of possibilities for sequence assembly. 

Starting with V2.4.0rc1 (corresponds to 2.3.31 of the development line), there
were no more restrictions built into the binary regarding time or number of
sequences that MIRA can handle. Your available memory will be the limit.

Starting with 2.4.0, binaries are now made available for both 32 and 64 bit
platforms of x86 Linux.

MIRA has learned a number of useful new tricks like assembling against
other sequences (backbones), usage of strain information in genomic
assembly (closely related strains can now be assembled in one go), SNP
analysis, optimised alignments (no more gap base jiggling), loading sequences
gained from the NCBI trace archive etc.

Compared to the 2.2.x line, speed has increased (sometimes quite
drastically) and memory requirements have decreased a bit. Several
smaller and bigger bugs have been fixed. I highly recommend to upgrade
to this version as soon as possible, even if parameters could not be
kept 100% backward compatible.

New / changed features
......................
- added possibility to load "backbones" and assemble against those
  sequences. Backbones can be in FASTA, CAF, EXP or even Genbank (GBF,
  GBK) format. Sequence features / tags are honoured in bankbones.
- enhanced inference of previously undetected repeat marker bases to
  include inference of IUPAC support.
- sequence alignment get nicer for "long" indel regions.
- alignment scoring function now per default assigns decreasing gap
  extension penalties. Eases life for assembling against genomic
  backbones. Drawback: -AL:egp must now be manually selected for
  for EST assembly and for "real" genome assembly, -CO:amgb:amgbemc
  are recommended.
- enhanced handlng of repetitive sequences characterised not by bases,
  but by insertions and deletions.
- enhanced contig tagging mechanism
- added counts of IUPAC and funny characters in contig statistics and
  _info_contigstats.txt files
- small change in output when parameter parsing failed: usage is now
  printed before analysis of error cause
- changed position columns in different _info and _out files so that
  now padded and unpadded positions are given.
- small cosmetic changes in different output files
- renamed FASTA output files: "raw" files are now named "padded" while
  previously 'normal' FASTA files without special extension are
  "unpadded". (getting some consistency with gap4)
- new result file type TCS: Transposed Contig Summary. Idea "borrowed"
  *cough* from TIGR .tcov files. Nicely suited for "quick" analyses from
  commandline tools or even visual inspection. Written only as final
  result with appendix "_out.tcs". New parameter: -OUT:ors
- first draft of SNP analysis function, saved in assembly information
  file "_info_snpanalysis.txt"
- Can now load GenBank files as backbone reads (new -SB:bft parameter
  value: "GBF"). Also load the features as GAP4 compatible tags from
  that file.
- larger changes in the tag naming scheme that also have repercussions
  in the parameter options. This was needed to simplify searches for
  problematic assemblies in editors (like e.g. gap4).
  Repeat Marker Bases (RMB) are now split into Strong/Weak types and
  also whether they occur in reads or in the consensus. PRMB becomes
  SRMr or SRMc, WRMB becomes WRMr or WRMc.
  The tags PAOS, PIOS and PROS for SNPs are now SAOr/c, SIOr/c and
  SROr/c.
  To keep parameter options naming scheme consistent, some parameters
  had to be renamed: -AS:pbl to -AS:rbl, and -CO:mpc:npz:mgqpt:mgqwpc
  to -CO:mrc:nrz:mgqrt:mgqwsc
- SRMc, SROc, SIOc and SAOc tags now get the group quality for each
  base as additional output in the "_info_consensustags.txt" file
- cleaned up "_info_consensustags.txt" and "_info_readtags.txt" a bit
- cleaned up error messages when SCF data is not found
- standard deviation of inserts are now read in NCBI traceinfo
  files. Minimum and maximum insert sizes are now calculated as
  insert_size -/+ 4*stddev.
- MIRA now automatically corrects sequence names in sequences
  downloaded from the NCBI traceinfo archive. It replaces the
  "gnl|ti|....." name with the "real" name (the one after the " name:"
  string). This allows using FASTA file from the trace archive
  directly without further preprocessing.
- if strain names are given, MIRA now also creates extra strain files in
  FASTA format as result of the assembly
- the parameters with which MIRA was called are now written at the start of a
  project into <projectname>_info_callparameters.txt

Performance
...........
- removed debugging code that was wrongly left activated in
  calculation of dynamic programming matrix (for alignments). Speedup
  in alignment calculation: factor ~6. Speedup in typical assembly
  project: factor >2. Bug was introduced in V2.2.3 (*sigh*)
- reduced memory consumption of sequences
- reduced memory consumption in assembly process: clipping of vectors
  (-CL:pvc) inflicted a huge memory penalty. This has been resolved.
- optimised assembly when loading backbones (rails are not aligned
  anymore)
- improved handling of similar sequences having (certain) indels:
  these are now treated as real indels. Takes effect when -CO:amgb is
  on.
- improved genome building anchors by starting in non-multicopy sites
- optimised skimming evaluation
- small internal speed optimisations
- speed enhancements: reads that have contradicting PRMBs will now be
  excluded from the SKIM and alignment phases in subsequent loops
- major speed increase (> factor 10) when loading larger CAF files

Tools
.....
- streamlined tools: merged several small utilities into 'scftools'
  and 'fastatools' 

Options
.......
- removed unused options: -AL:emp* (not used since a long time),
  -CO:mgqwsc:nrz (disappeared in 2.3.29)
- renamed -CO:ismin:ismax to -CO:dismin:dismax
- added -OUT:ots for tcs output of temporary results
- added -CO:np
- -CO:asir is now also setable by commandline (was reserved for setting
  only by miraEST)
- added -SB: options. Also -AL:megpp -CO:amgb:amgbemc
- renamed -GE:ess:lb:bft:brl parameters to new category -SB
- renamed -EG:ess to -GE:ess. Moved -EG:lsd to -SB:lsd (and disbanded -EG category)
- added -AS:sel -SK:mhpr to optimize performance for really deep
  repeats. Only "n" best hits are given to the SW alignment checks
- added logic to improve assembly when some reads have too high quality values
  for wrongly called bases
- added quick switches -estmode and -horrid
- quick switches on the command line now print out what they are setting

Bugfixes
........
- base positions now won't get multiple equal tags.
- SNPs were sometimes wrongly disrupting contig building
- -CO:amgbemc was (still) not honoured. Fixed.
- the problem of slow loading EXP files has been resolved
- tags lost their direction when saved as CAF or ACE, fixed.
- tags got wrong direction when loaded from CAF
- quality in CAF files were put in one single large line,
  fixed to multiline
- -CO:also_mark_gap_bases and -CO:also_mark_gap_bases_even_multicolumn 
  did not work as advertised
- several small bugfixes in output functions that led MIRA to abort on
  rare occasions while saving results to files.
- CAF and ACE files now get "correct" multiline tags
- TG tags in EXP files were all converted to be on both directions
- fixed small bug while parsing -SB parameters
- fixed ugly bug in template handling (introduced 2.3.11, the 2.2.x
  line was fortunately not affected). This led to really bad
  assemblies when template information was used.
- potential problem fix: changed output in EXP files for ON entries to
  be now multiline, so that the Staden iolib can cope with large
  entries
- the unpadded fasta quality result file contained, in fact, the
  padded fasta results, fixed.
- in rare cases, low quality bases were taken into account when
  searching for Possible Repeat Marker Bases (PRMBs). Fixed.
- in rare cases, mira would stop in loops>1 when internal tag
  handling discovered an error. Error cause has been fixed.
- skimmer: some hits for non-exact matches were not found.
- CAF files containing "Ligation_no" lines caused errors while reading
  them
- HTML output function crashed on some systems, this should now not happen
  anymore. 
- progress counter during contig building makes "nicer" progress
  report
- several small typos: parameter options should now be in sync again
  with the documentation (man pages etc.)




Changes in detail:
==================

2.6.0rc1 to 2.6.0rc2:
---------------------
- improved contig building time by reducing need for alignment recalculation
- fixed rare problem that led to abortion of mira in contig building

2.5.12 to 2.6.0rc1:
-------------------
- introduced DWIM (Do What I Mean) parameter switches:
    -genomedraft, -genomenormal, -genomeaccurate
    -mappingdraft, -mappingnormal, -mappingaccurate
    -clippinglight, -clippingnormal, -clippingheavy
    -highlyrepetitive
    -highqualitydata
- fixed dumb error in vector clipping (-CL:pvc) that clipped away too much
  when vector clipping was performed on a read
- fixed rare problem where mira aborted while extending reads. (-DP:ure)
- optimised partioned skim parameters for present day quality Sanger type
  shotgun sequences. Runtime -18%. No effect with 454 type data.
- fixed errors in parameter parsing: -OUT:org:orf -AS:sel
- quick switches for loading files (e.g. -fasta, -phd etc.) have been cut back
  to exactly this functionality (loading), without further side effects (like
  switching of read extension etc.). Functionality transfered to DWIM switches
  (see above)

2.5.12
------
- improved algorithm for computing read extensions on Smith-Waterman aligned
  sub-sequences, now switched on per default for pre-assembly read extension
- added -DP:rewl:rewme:feil:leil
- added -CO:amgbnbs

2.5.11
------
- fixed a problem while loading CAF files with exposed sequence vectors
  that led to wrong sequence positioning in a contig
- when CAF files loaded as backbone do not contain contigs, then the reads
  themselves are used as backbones.
- added special mode for backbone assembly, faster and more accurate
- changed genomic assembly mode, faster and more accurate

2.5.10
------
- deleted -CO:mrc parameter
- added -CO:mrpg parameter

2.5.9
-----
- reworking of internal data structures leads to 37% decrease of memory
  consumption of stored alignment data. Important for 454 type data,
  e.g. project with 250000 reads now uses 850M for this part instead of
  1350M.
- improved alignment algorithms lead to better alignments of 454 type data
  (improvements also noticable for Sanger type data, but less so)
- bugfix: in very rare cases, sequences could be wrongly inserted into a
  consensus, leading to slight misalignments. Problem seen first with 454 type
  data
- new parameter: -SB:abnc

2.5.8
-----
- railreads and backbones in contigs now have no SCF files (and exp) 
  files assigned
- new parameters: -454:mdis454:hybrid (partly functional)

2.5.7
-----
- renamed -CO:uti to -GE:uti
- MIRA switches off usage of template information when no useful information
  for this is present in data
- changes in pathfinder to speed up building contigs with many reads (454
  data) or when genome sized backbones are used
- new "hidden" parameter: -PF:swcs
- CAF sequence names may now contain "#" and "|" as sequence names (the later
  perhaps not being very useful)
- FASTA sequence names may now contain "#" as character
- FASTA sequence names containing "|" are rewritten to contain "_"

2.5.6
-----
- added additional Genbank tags as "dont analyse for SNPS" in
  assout::saveFeatureAnalysis()
- first test versions capable of acceptable 454 data assembly

2.5.5
-----
- adjusted typos in documentation: -CO:dismin and dismax lacked the "d" in
  some parts of the docs (due to a parameter rename earlier *sigh*)
- fixed typos in -CL:mlcr:smlc documentation

2.5.4
-----
- new partitioned skim routine: faster, less memory consumption. Current
  drawback: number of hits cannot currently be limited per read (-SK:mhpr has
  no effect)
- new quick switch: -454data. Please note that as of this writing, mira is not
  yet optimally suited to handle this type of data well, there is still some
  development needed in this area. Relying on the consensus of MIRA is NOT
  recommended at this time!
- ommission in manual corrected: -CL:emlc:mlcr:slmc were not described.

2.5.3
-----
- GBF output now complete: protein translation also written when Genbank
  features are present as tags in the assembly
- convert_project: fasta output now also writes out the consensus of contigs
  in an assembly when the input was CAF
- convert_project: added -q
- convert_project: added aliases caf2fasta, caf2gbf, caf2text, caf2html,
  gbf2caf and gbf2fasta as aliases to convert_project which have -f and -t
  already set accordingly

2.5.2
-----
- decreased memory usage for skim routine by 50%, almost non-noticable speed
  penalty

2.5.0 ->
--------
- ace2caf: new option -F and small improvements
- small bugfixes
- first GBF output


2.4.0rc2h
---------
- added -SB:sbuil parameter option
- fixed bug when reading GBF files where positions have a ">" in front
- change: assembly of multiple strains improved
- change: /note entries are now also read from GBF files and put into tags
- change: tags for Repeat Marker Bases and SNPs in reads now do not get any
  comment, only the consensus tag gets the full comment. Saves quite some
  space in .exp and .caf files for projects with many such tags. 
- added GBF as outtype for the "convert_project" program
- added demodata for backbone assembly


2.4.0rc2c to 2.4.0rc2g
----------------------
- added logic to improve assembly when some reads have too high quality values
  for wrongly called bases
- added quick switches -estmode and -horrid
- quick switches on the command line now print out what they are setting
- small bugfix in determination of multicopy reads when using backbone
  assembly
- internal rearrangements


2.4.0rc2 to 2.4.0rc2c
---------------------
- the parameters with which MIRA was called are now written at the start of a
  project into <projectname>_info_callparameters.txt
- improved pathfinder that now better bridges repetitive sequences
- fixed bug that struck the pathfinder (endless loop) when a) disk was full
  and b) -AS:max_contig_buildtime was used
- major speed increase in pathfinder when using backbone assembly. Rearranged
  pruning in pathfinder, leading to dropping untaken paths earlier in the
  evaluation. Results of the pathfinder are the same as in earlier versions,
  no change there. 
- improvement in handling of repeats characterised only by indels: these are
  now correctly treated as are repeats characterised by basechanges. leads to
  vastly improved assemblies of repetitive regions. Also due to SRMx tags that
  were triggered by gaps now induce multibase (multicolumn) tags. These are
  also set on the ends of stretches of equal bases
- backbone contigs consisting of a single sequence now get the name of the
  sequence as name of the contig 
- if strain names are given, MIRA now also creates extra strain files in
  FASTA format as result of the assembly
- major speed increase (> factor 10) when loading larger CAF files


2.4.0rc2
--------
Mostly bugfixes and last tweaking on small things. This should be the
last release candidate before 2.4.0, which will include more
documentation and examples.

- bugfix: a fatal bug in 2.4.0rc1 prevented traceinfo XML files from
  being correctly read. So, when one relied on the XML files, this
  lead to sequencing vector not being clipped, bad quality being
  included etc. This has been fixed.
- bugfix: HTML output function crashed on some systems, this should
  now not happen anymore.
- progress counter during contig building makes "nicer" progress
  report
- added documentation entry for -AS:ugpf
- standard deviation of inserts are now read in NCBI traceinfo
  files. Minimum and maximum insert sizes are now calculated as
  insert_size -/+ 4*stddev.
- MIRA now automatically corrects sequence names in sequences
  downloaded from the NCBI traceinfo archive. It replaces the
  "gnl|ti|....." name with the "real" name (the one after the " name:"
  string). This allows using FASTA file from the trace archive
  directly without further preprocessing.


2.3.30
------
- removed unused options: -AL:emp* (not used since a long time),
  -CO:mgqwsc:nrz (disappeared in 2.3.29)
- enhanced inference of previously undetected repeat marker bases to
  include inference of IUPAC support.
- removed debugging code that was wrongly left activated in
  calculation of dynamic programming matrix (for alignments). Speedup
  in alignment calculation: factor ~6. Speedup in typical assembly
  project: factor >2. Bug was introduced in V2.2.3 (*sigh*)

2.3.29
------
- alignment scoring function now per default assigns decreasing gap
  extension penalties. Eases life for assembling against genomic
  backbones. Drawback: -AL:egp must now be manually selected for
  for EST assembly and for "real" genome assembly, -CO:amgb:amgbemc
  are recommended.
- sequence alignment get nicer for "long" indel regions.
- enhanced handlng of repetitive sequences characterised not by bases,
  but by insertions and deletions.
- enhanced contig tagging mechanism
- bugfix: base positions now won't get multiple equal tags.

2.3.28
------
- further working on SNP/Feature analysis
- bugfix: SNPs were sometimes wrongly disrupting contig building
- streamlined tools: merged several small utilities into 'scftools'
  and 'fastatools' 
- renamed -CO:ismin:ismax to -CO:dismin:dismax
- bufix: -CO:amgbemc was (still) not honoured. Fixed.

2.3.27
------
- further working on SNP/Feature analysis
- bugfix: the problem of slow loading EXP files has been resolved
- bugfix: tags lost their direction when saved as CAF or ACE, fixed.
- bugfix: tags got wrong direction when loaded from CAF
- first internal changes to make compiling -Wall ... etc. proof

2.3.26
------
- bugfix: quality in CAF files were put in one single large line,
  fixed to multiline
- multiple internal changes (char * to string, shifting of functions into
  namespaces etc.)

2.3.25
------
- bugfix: -CO:also_mark_gap_bases and -CO:also_mark_gap_bases_even_multicolumn 
  did not work as advertised
- reduced memory consumption of sequences
- reduced memory consumption in assembly process: clipping of vectors
  (-CL:pvc) inflicted a huge memory penalty. This has been resolved.

2.3.24
------
- several small bugfixes in output functions that led MIRA to abort on
  rare occasions while saving results to files.
- added counts of IUPAC and funny characters in _info_contigstats.txt
  files

2.3.23
------
- small change in output when parameter parsing failed: usage is now
  printed before analysis of error cause
- changed position columns in different _info and _out files so that
  now padded and unpadded positions are given.
- small cosmetic changes in different output files

2.3.22
------
- added -OUT:ots for tcs output of temporary results
- fleshed out the SNP analysis
- renamed FASTA output files: "raw" files are now named "padded" while
  previously 'normal' FASTA files without special extension are
  "unpadded". (getting some consistency with gap4)

2.3.21
------
- bugfix: CAF and ACE files now get "correct" multiline tags
- new result file type TCS: Transposed Contig Summary. Idea "borrowed"
  *cough* from TIGR .tcov files. Nicely suited for "quick" analyses from
  commandline tools or even visual inspection. Written only as final
  result with appendix "_out.tcs". New parameter: -OUT:ors
- first draft of SNP analysis function, saved in assembly information
  file "_info_snpanalysis.txt"

2.3.20
------
- bugfix: TG tags in EXP files were all converted to be on both
  directions
- Can now load GenBank files as backbone reads (new -SB:bft parameter
  value: "GBF"). Also load the features as GAP4 compatible tags from
  that file.
- optimised assembly when loading backbones (rails are not aligned anymore)

2.3.19
------
- larger changes in the tag naming scheme that also have repercussions
  in the parameter options. This was needed to simplify searches for
  problematic assemblies in editors (like e.g. gap4).
  Repeat Marker Bases (RMB) are now split into Strong/Weak types and
  also whether they occur in reads or in the consensus. PRMB becomes
  SRMr or SRMc, WRMB becomes WRMr or WRMc.
  The tags PAOS, PIOS and PROS for SNPs are now SAOr/c, SIOr/c and
  SROr/c.
  To keep parameter options naming scheme consistent, some parameters
  had to be renamed: -AS:pbl to -AS:rbl, and -CO:mpc:npz:mgqpt:mgqwpc
  to -CO:mrc:nrz:mgqrt:mgqwsc
- the -SB:bol option has been removed

2.3.18
------
- fixed small bug while parsing -SB parameters
- improved handling of similar sequences having (certain) indels:
  these are now treated as real indels. Takes effect when -CO:amgb is
  on.

2.3.17
------
- fixed ugly bug in template handling (introduced 2.3.11, the 2.2.x
  line was fortunately not affected). This led to really bad
  assemblies when template information was used.
- improved genome building anchors by starting in non-multicopy sites
- new utility program: scf_remix. Useful for "fixing" broken SCFs or
  SCFs that are out of sync with other data sources

2.3.16
------
- added -CO:np
- -CO:asir is now also setable by commandline (was reserved for setting
  only by miraEST)
- PRMB, PROS, PIOS and PAOS tags now get the group quality for each
  base as additional output in the "_info_consensustags.txt" file
- cleaned up "_info_consensustags.txt" and "_info_readtags.txt" a bit
- added counts for IUPAC bases and funny characters in contig
  statistics

2.3.15
------
- added -SB:bol:bbq -AL:megpp -CO:amgb:amgbemc
- cleaned up error messages when SCF data is not found
- potential problem fix: changed output in EXP files for ON entries to
  be now multiline, so that the Staden iolib can cope with large
  entries


2.3.14
------
- renamed -GE:ess:lb:bft:brl parameters to new category -SB
- added possibility to load FASTA files as backbone (-SB:bft=FASTA)
- added possibility to give backbone sequences a strain name (-SB:bn)
- bugfix: the fasta quality result file contained, in fact, the raw
  fasta results, fixed.
- bugfix: in rare cases, low quality bases were taken into account
  when searching for Possible Repeat Marker Bases (PRMBs). Fixed.


2.3.13
------
- added possibility to load "backbones" (CAF) and assemble against
  those. New parameters -GE:lb:bft:brl
- renamed -EG:ess:lsd to -GE:ess:lsd (and disbanded -EG category)
- bugfix: in rare cases, mira would stop in loops>1 when internal tag
  handling discovered an error. Error cause has been fixed.


2.3.12
------
- bugfix in skimmer: some hits for non-exact matches were not found.
- optimised skimming evaluation
- added -AS:sel -SK:mhpr to optimize performance for really deep
  repeats. Only "n" best hits are given to the SW alignment checks
- small internal speed optimisations


2.3.11
------
- small bugfix: CAF files containing "Ligation_no" lines caused errors
  while reading them
- speed enhancements: reads that have contradicting PRMBs will now be
  excluded from the SKIM and alignment phases in subsequent loops
- several small typos: parameter options should now be in sync again
  with the documentation (man pages etc.)

2.2.7 to 2.2.8 (2.3.10)
-----------------------

This is an intermediate optimisation release. Although I wanted to
build in some more (exiting) new features, especially the spoiler
detection and the speed improvements justify releasing the
improvements as they are. They make assembly of genomes from 1 to 10mb
a bit more fun.

- added -AS:sd:sdllo to detect and remedy assembly "spoiler". Only
  recommended for assembly of genomic sequences! These spoiler can be
  either chimeric reads or reads with long parts of unclipped vector
  sequence (that was too long for the -CL: vector leftover
  clippings). These spoiler typically prevent contigs to be joined,
  MIRA will cut them back so that they present no more harm.
- added -GE:rns to support naming schemes of different sequencing
  centers. Sanger and TIGR naming schemes are now supported.
- added -GE:pd flag for controlling date output in ouput log
- major speed improvements for projects where large contigs (>500kb)
  with many reads are built.
- minor bugfix: some overlap where not correctly recognised by the
  SKIMmer
- minor bugfix: now got the percentage progress report bar right, it
  sometimes showed false status. 
- minor bugfix: added -AS:umcbt:bts to adapt for larger assemblies on
  slower machines. More useful for EST assembly than for genomic.
- starting with gcc 3.4, mira is now compiled with -O3 as standard
  optimisation level. Gain of ~5-10% in many algorithms
- known issue: I apparently "optimised" some pathfinding routines too
  much for EST data. Sometimes, for genomic data, some contigs are not
  at their optimal length. This most likely occurs in low coverage
  shotguns (<=4), high coverage (>=6) should be ok. I'm working on it.

2.2.6 to 2.2.7 (2.3.9)
----------------------
- fixed serious bug that led to suboptimal assembly of genomic
  sequences. Upgrade to this version *highly* recomended.
- added -OUT:oetas, exttmp singlets now not saved by default 
- exttemp contigs now not saved as "post" when no change (either
  repeat marked or edits) happened
- *sigh* fixed error in parameter setting for extended_gap_penalty:
  _long_ gaps sometimes were given a lower penalty than expected
- fixed error in .txt output of a contig: some HTML was thrown in
  sometimes

2.2.5 to 2.2.6
--------------
- fixed a bug in .ace files that prevented consed to load them
  ("clview" from TIGR was not affected)
- fixed rare bug that led to an assembler panic and subsequent abort 
  of the assembly process.
- brought the provided demonstration parameter file up-to-date

2.2.4 to 2.2.5
--------------
- fixed typo in miraEST internal standard parameters, miraEST would not
  start

2.2.3 to 2.2.4 (2.3.7)
----------------------
- added ability to merge data from NCBI trace info files in XML format
  (-GE:mxti and -FI:xtii)
- put -GE:lsd:ess to new group -ESTGENERAL (-EG:lsd:ess)
- general update and overhaul of documention

2.2.2 to 2.2.3
--------------
- fixed typos again in on-screen text

2.2.1 to 2.2.2 (2.3.3)
----------------------
- fixed ugly bug in banded Smith-Waterman that lead to misses in some
  cases

2.2.0 to 2.2.1
--------------
- fixed typos in on-screen text

2.1.22 declared 2.2.0, forked 2.3 branch
----------------------------------------

2.1.21 to 2.1.22
----------------
- fixed minor bug in computation of alignment scores. It lead
  sometimes to suboptimal alignments at the ends of an overlap. Effect
  was frequently seen in EST projects
- new parameters -AL:extra_mlsmatch_penalty:emp*
- new FASTA output for the consensus: the raw format, with gaps,
  lowercase for normal consensus, upper case for special features like
  PRMB, WRMB, PAOS, PROS, PIOS
  files are named .raw.fasta

2.1.20 to 2.1.21
----------------
- miraEST gets starting step as parameter: -GE:ess
- fine tuning of gap penalty level for est_splitsplices variant

2.1.19 to 2.1.20
----------------
- optimised memory requirements when reading FASTA files
- slight optimisation of memory requirements for reads
- fixed bug when reading FASTA quality files that had more sequences
  than the FASTA files themselves
- fixed a number of minor internal bugs that were found with valgrind
  which had no traceable effects on the assembly
- introduced the TEST versions of MIRA and miraEST

2.1.18 to 2.1.19
----------------
- fixed small bug that caused some reads with PRMB/WRMB tags that
   matched ok to be rejected as overlap. Influence on assemblies:
   light, but annoying

2.1.17 to 2.1.18
----------------
- fixed dumb bug in SKIM which caused suboptimal hit numbers
  *deepsigh*
- fixed small memory leaks here and there (valgrind rocks)


2.1.16 to 2.1.17
----------------
- fixed (dumb dumb dumb) bug: check of minimum alignment score did not
  take the score multiplier into account *sigh* this resulted in good,
  but somewhat shorter alignments to be rejected 
- slight tweak in consensus algorithm (less IUPAC codes)
- new parameter: -DP:pvcmla. Enables quite effective sequencing vector
  leftover clipping without loosing splice variants (variants with
  lower number of bases than -DP:pvcmla will get lost though).
- Sequences not adhering to Sanger (and probably St. louis) naming
  scheme now loose template information. Allows better assembly for
  projects that don't have this scheme.
- miraEST comes with some enhanced standard parametersets

2.1.15 to 2.1.16
----------------
- changed behaviour of contig: when assume_snp_instead_prmb, now also
   tags PROS as PRMB
- polybase masking now uses an enhanced algorithm, -DP: options
   changed too to reflect this
- minor enhancements in IUPAC consensus computation

2.1.14 to 2.1.15
----------------
- Reading of SCF V2 was borken on x86, fixed

2.1.13 to 2.1.14
----------------
- fixed bug that affected consensus: in (really) rare cases, a base in
   the consensus was replaced by another base in the consensus output

2.1.12 to 2.1.13
----------------
- improved consensus quality calculation: supporting reads add a bit
   more to a quality
- fixed dumb bug *sigh* that lead to suboptimal assembly results in
   some cases involving PRMB tags
- during assembly, contigs are now only edited when no unresolved
   misassembly was detected in that contig. (TODO: auch bei nur WRMBs?)

2.1.11 to 2.1.12
----------------
- added POLY and IUPAC tags for HTML output
- small bugfix in HTML output for MISM tag
- bugfix in HTML output: tags in consensus are now shown
- progressbar when loading FASTA files re-enabled
- miraEST now names single-read-contigs (result of step 2) now
   _Singlet instead of _Contig
- helper programs (scf2other etc.) now mention the MIRALIB version in
   their usage text

2.1.10 to 2.1.11
----------------
- adapted base version for diss

2.1.9 to 2.1.10
---------------
- tweaked consensus base probabilities
- marking of repeats: only when dubious bases are surrounded by good
quality (new parameter -CO:mnq)

2.1.8 to 2.1.9
--------------
- improved consensus algorithm when non-clipped vector leftovers occur
- improved tagging of possibly misassembled repeats: single read
misassemblies now better under control
- fixed off by one bugs in tagging of polybases at read ends
- new parameter class -SKIM

2.1.7 to 2.1.8
--------------
- improved consensus algorithm for uncertain base/gap candidates
- renamed -AL:gpl=est_default to est_splitsplices
- new parameter class -DATAPROCESSING
- moved -AS:mr put to -CO:mr and -AS:ure to -DP:ure
- new parameter option -DP:tpae to enable/disable tagging of poly-A/T
  at read ends, options for polybase tagging in DP

2.1.2 through 2.1.7
-------------------
- Consensus disregards bases that are 'masked from consensus' (for the
time being the tag POLY for poly-A or poly-T at ends of reads)
- Consensus is now given with IUPAC bases if base evidence is
  contradictory
- Overall improved IUPAC support
- bugfix: clusters were wrongly computed (affected only output)
- SKIM algorithm can be parameterised
- new extra gap penalty level (egp): 10 (est_default)


2.0.1 to 2.1.2
--------------
- bugfix: parameters for clippings (qual and masked chars) were not used (only 'defaults')
- added computing and output of possible clusters
- better alignments for difficult cases
- writes clustering logfiles
- emergency search stops now work on time dependent basis
- reworked blacklisting of reads for EST assemblies 

2.0.0 to 2.0.1
--------------
- fixed typos
- -AS:uess:esspd for restraining computing time on pathological cases
  of coverages are now functional
- fixed bug, standard parameters for third step of miraEST were in
  wrong section














 - Fixed error in contig: wrong assumption about insert sizes lead to
   halt.
 - New Pathfinder algorithm (faster in resolving) CHECKME!
 - New banning strategy for found misassemblies CHECKME!
 - New parameter -CO:emea, -DI:gap4da
 - Added .ace output (alpha) (Tags?)
 - Added gap4 directed assembly output 
 - Screened out bases (Xs) are now (durchgeschleift) and not
   transformed to Ns in reads anymore.
 - Sequences loaded as FASTA can now also fall back on SCF files for
   qualities and editing if those are present
 - standard filenames for in and out changed to "mira"
 - fixed bug in EdIt that caused crashes when SCF was not present
   (thank you valgrind! :).
 - added -GE:project to quickly change standard filenames for in and
   out.
 - added quickparams --fasta, --project, --phd
 - quietened EdIt when analysing stretches containing reads with no
   SCF data
 - switching automatic contig editing off when no SCF present for the
   reads
 - added possibility to save contig consensus (and qualities) as
   FASTA: parameters -GE:orf:otf
 - Singlets are now named "Singlet" instead of "Contig" in result
   files. They still get the same continuous number as if they were
   contigs though.
 - contigs are now more permissive on errors when template partners
   are in range (rodirs*2)
 - added -CO:ismin:ismax for controlling default template insert size
 - PHD files can now be read, added -FI:pi:fpi (fofnphd tut noch nicht)
 - relocated output parameters to -OUTPUT, added extended temporary output flags 
 - when loading from fasta or phd, template names are now deduced
   from read name if they're in Sanger Centre scheme
 - added options to perform clipping on reads by quality
   (-CL:qc:qcmq:qcwl)
 - added options to perform clipping on reads by masked bases
   (-CL:mbc:mbcgs:mbcmfg:mbcmeg)
 - repositioned -GE:cpv to -CL:pvc
 - added _reads_invalid and _reads_too_short as output files
 - added several info and error files as output (for statistics etc.)
 - cleaned output as text and put into file
 - speeded up read extension
 - SCF files are now found even if the filenames differ from given
   names by appending a .Z or .gz, .scf, .scf.Z etc.
 - practically doubled speed of banded SW alignment using memrecache
   algorithms (yeeehaah!) 
 - added -CO:mgcpt:mgcwpc
 - added possibility to load parameters from file (-params)
 - added -GE:discard_read_on_eq_error
 - write a lot of statistics files at the end of an assembly
 - added clustering log files

TODO: Statistiken und Listen (Orphans / Singlets / Cluster?)
TODO: fofnphd realisieren
TODO: mehrere Inputfiles laden, vorher anzahl reads zhlen und per reserve() anmelden
TODO: singlets in andere files trennen?
TODO: contigs.C: template handling in addRead verbessern
TODO: versteckte STL Containerspeicherlecks (ungengende reserve()) suchen

*Changes from V1.5.2 to V1.5.3
 - Added -GE:cpv for clipping possible sequencing vector leftovers in
   reads (just on the left side at the moment)

*Changes from V1.5.1 to V1.5.2
 - Added new parameter -DIR:exp:scf:log to specify input and output
   directories (log doesn't work yet)
 - Fixed a bug that caused segmentation faults when SCF files with 0
   bases were used.

*Changes from V1.5 to V1.5.1
 - Added new parameters -AS:pbl (maximum prmb break loops)  -CO:npz (num_prmb_zones)
 - new function to transfer sequencing vectors expressed as tags in
   EXPs to clips: searches with a tolerance from clips and strat/end
   of read, transfers tags found there to clips.
 - New! EST assembly now supported by usage of strains. Added new
   parameters -GE:lsd and -FI:sdi
 - New routine for finding possible repeats (PRMB) and possible SNPs (PSNP)
 - Added WRMB as weak repeat marker bases
 - Comments are now allowed in the file of filenames file (fofn)

*Changes from V1.4.1  to V1.5
 - New read comparison routine (experimental): Skim. Speed factor to
   the Zebra routines: 10 to 50, depending on memory. Drawbacks:
   probably isn't as sensitive as ZEBRA, no possibility to subdivide
   the search space for the time being.
 - Change in behaviour while loading SCF files: 'fatal' errors in SCFs
   now do not lead to a halt, but are logged (and the reads concerned
   excluded from the assembly).
 - option -FI:fastaqualin added. new FASTA reading routines now load
   also quality files in FASTA format.
 - fixed template handling bug: distance of reads was calculated
   wrongly in the contig (affected assembly only when -CO:uti was on).
 - New building mechanism using automake and autoconf.

*Changes from V1.4.0rc2 to V1.4.1
 - the option to load FASTA files was lost somewhere in earlier
   revisions, thanks to the people who pointed that out.

*Changes from V1.4.0rc1 to V1.4.0rc2
 - option -GE:filecheck_only added
 - logfile "log.scfread_fail" added
 - Bug in EdIt removed that sometimes caused crashes on unclipped
   sequences 
 - Bug in parameter parsing removed that caused wrong parameters not
   to be recognised.

*Changes from V1.3.20 to V1.4.0rc1
 - Reworked HTML format a bit.
 - Merged tools for project conversion into convert_project.
 - MIRA complained when it encountered PHRED SCF files that contained
   irregularities/errors. It will now 'correct' the error internally
   and continue.
 - Read extension is now additionally performed _before_ the first
   assembly (if read extension is enabled)
 - new command line option '-borg'. This will trigger a lot of
   parameters to be set into a mode where MIRA is likely to assemble
   everything that might look like ok to assemble. Albeit this slows
   down the assemble _a lot_.
 - Integrated editor had a few bugs fixed.
 - A few code cleanups

*Changes from V1.3.19 to V1.3.20 (maintenance release)
 - The EdIt routines for ALF data (mira_l) had not been actualised and
   were not working right: a lot of bases that could have been
   corrected were not corrected.
 - In rare cases, buggy SCF files caused the integrated editor to
   crash. Fixed by augmenting the ability to recognise buggy SCFs.
 - Added optional HTML output for contigs
 - Added -GE:orh, -GE:otc and -GE:oth (see man page) for controlling
   html output and temporary CAF|HTML output files

*Changes from V1.3.18 to V1.3.19
 - Added -FILE options. File and project names can now be freely chosen 
 - CAF read routine had a small bug. Affected people who worked from
   the very first base in reads (that wasn't clipped off through
   quality and/or sequencing vector)

*Changes from V1.3.17 to V1.3.18
 - Small changes in EdIt. Bugfixes in MIRA and EdIt. If MIRA didn't
   crash on you, you probably weren't affected.

*Changes from V1.3.16 to V1.3.17
 - Argl, bad bug in 1.3.16 while loading files which caused mira to
   crash. Sorry.

*Changes from V1.3.15 to V1.3.16
 - Integrated EdIt had memory leak and crashed
 - Minor bug in CAF writing fixed for Solaris and Linux version. Bug
   did not affect quality of assembly, it's just that not-existing
   Clonevec names were replaced by the string "(null)".

*Changes from V1.3.14 to V1.3.15
 - The EdIt (automatic editor) routines contained an error that struck
   in very rare cases and crashed the assembler when -GE:ace was
   set. Fixed. 
 - Memory requirements decreased again
 - Small bugfixes
 - SGI version now runs in true 64 bit mode

*Changes from V1.3.13 to V1.3.14
 - Fixed error in handling of repeat marker bases (this could have
   lead to a crash)

*Changes from V1.3.12 to V1.3.13
 - Substantially decreased memory requirements phase. Well,
   requirements decreased incredibly, drastically, dramatically,
   ... you get the picture.
 - Fixed small bug when compiling with gcc: filenames were sometimes
   garbled (Linux, Solaris)

*Changes from V1.3.11 to V1.3.12
 - added extended checkpointing: each contig is now saved separately during
   the assembly in log.loop_W_cbX_iY_Z.caf where W ist numeric loop
   number, X is numeric contig number in this loop, Y is numeric
   iteration number for this contig in this loop, Z ist either 'pre'
   or 'post' - indicating before the contig has been edited or after.
 - Inserted or changed bases now get a quality value != 0. The quality
   is interpolated from neighbouring non-N and non-gap bases. Rough,
   but works.

*Changes from V1.3.10 to V1.3.11
 - Added checkpointing capability (files: bla_out_loop.X.caf) where X
   stands for the loop number.
 - Fixed severe bug that caused MIRA to stop. Introduced somewhere in
   1.3.x 

*Changes from V1.3.9 to V1.3.10
 - First prototype auf automatic repeat marker

*Changes from V1.3.8 to V1.3.9
 - Added -AS:nol, -AS:ure and -AS:ace
 - Added -CO:uti to try making use of template information (insert
   size)

*Changes from V1.2 through V1.3.8
 - First integration of MIRA with EdIt, the automatic editor.
 - Added template handling
 - Added consensus tags
 - Removed memory leaks
 - bugfixes
 - some more bugfixes
 - tons of bugfixes *sigh* (note: if the program did not stop or
   crash in previous version, you were NOT affected, all of your
   assemblies were correct)

*Changes from V1.1.1 to V1.2
 - Added -AL:bip, -AL:bmin and -AL:bmax options, to make banded SW
   configurable

*Changes from V1.0.1 to V1.1.1
 - Now using banded Smith-Waterman alignment functions. Speed increase
   between 300% and 700% in the alignment phases. BSW functions might 
   miss a valid alignment, but only in very very rare cases. BSW were 
   needed as labs increasingly show up with read length between 400
   and 1000 bases.
 - added IUPAC uncertainty codes for EXP, SCF and CAF reading
   routines. These will be treated as N internally and appear as N in
   the resulting alignment.
 - fixed bug: the signal analysis routines were never called
   (oooops). This bug appeared probably in 0.99b6.
 - fixed bug: temporary files in the SCF load function were not
   removed when an error occured

*Changes from V1.0 to V1.0.1
 - added -AL:mo parameter
 - version schemes now similar to the linux kernel. Even major numbers
   represent 'stable' version, uneven are 'test' version with features
   that weren't tested thoroughly on real data sets.

*Changes from V0.99b7 to V1.0 (not publicly released)
 - Faster filter functions with increased sensitivity and specificity
   built in. Filtering is now done with Zebra-Blocking instead of
   DNASAND. This is a major improvement in terms of speed (roughly 4x)
   in the filter phase.
 - Memory consumption in the assembly phase has been significantly
   reduced. It should now be perfectly possible to assemble projects
   with 50,000 to 100,000 reads (though _PLEASE_ contact the author
   before doing this, so that tips in speed enhancement can be given).
 - removed SANDSIEVE parameter options
 - added -AL:egp and gpl options
 - added ZEBRABLOCKING options
 - Used parameters are now dumped to stdout when MIRA starts
 - Unknown identifiers in EXP files do not generate warnings anymore
 - fixed reported bugs

*Changes from V0.99b6 to V0.99b7:
 - experiment files now don't need to be available when loading CAF
   projects
 - bug fixed in parsing command line options: -GE:lj=FOFNEXP wasn't recognised
 - some debug output removed that happened to be printed when loading
   CAF files
 - bugfix when reading experiment files: one line tags like
        TG   WARN - 127..167 "POSSIBLY VECTOR: puc18 289 249 2686"
   were misinterpreted
 - can now read quality values in EXP files
 - added -GE:eq and -GE:eqo options to specify quality sources

*Changes from V0.99b5 to V0.99b6:
 - bug fixed in CAF loading routines
 - potential bug fixed in contig handling
 - fixed bug in parsing command line options introduced in 0.99b3

Changes from V0.99b4 to V0.99b5:
 - changed logic for analysing danger zones (ALUS and REPT): checking
   should be stricter now
 - fixed bug in contig: in some rare cases, a division by zero error occured

Changes from V0.99b3 to V0.99b4:
 - fixed bug in dynamic programing algorithm: * in reads are now
   treated like N

Changes from V0.99b2 to V0.99b3:
 - switched on (experimental) possibility to reassemble CAF projects
 - added -GENERAL parameter options
 - temporary files are now removed automatically after the
   assembly. Use -GENERAL:clean_tmp_files=off if you plan to
   experiment with different assembly options.
 - fixed a bug: the -CONTIG:rej_on_dropinrelscore given as parameter
   was ignored
 - the -CONTIG:rej_on_dropinrelscore default is 7%, not 5 as I wrote
