HISTORY_genblast

/******************************************/
GenBlast Package Release CHANGELOG
/******************************************/

V101: Changes from genBlastG_070209: (July 3, 2009)
- change the program name to genBlast and integrate genBlastA and genBlastG into the "-p" option in genBlast;
- update README accordingly;

V102:
- typo fixed;

V103: (Aug. 9, 2009)
- added: check for existing "*.wublast" or "*.wublast.report" file 
("*" is the concatenation of query filename and target filename), 
if they already exist, then skip wublast or report-creating step;
- fixed: wublast format file parsing fixed (to parse wublast result for one query);

V104:
- input pathname fixed;

V105:
- problem with multi-line query gene name in wublast file fixed;

V106:
- problem with empty list of HSP in multi-query wublast file fixed;
- output ID naming duplicates problem fixed;

V107:
- fixed problem with empty target string when extending alignment during exon-repairing;

V108: (Aug. 14, 2009)
- fixed gff output so gene start/end is the same as exons after repair;
- make sure gff exons always start from smaller coordinate;
- fixed output rank counting;

V109:
- fixed problem with heading/trailing stop codon in alignments after overlap resolving

V110:
- outputs for -norepair option updated

V111:
- some more tweaks to the algorithm:
* for "no-intron" case, the donor-side and acceptor-side HSPs use the same as the current 
top-ranked pair of donor-acceptor
* if there's in-frame stop within HSP, always make that an intron region
* if two neighboring HSPs (after overlap resolving) are not in same frame, always assume 
the region between is intron region
* when computing exons for each intron region, only consider "no-intron" case when there's
no in-frame stop in the region, also when the donor-side and acceptor-side HSPs are in 
same frame
* if the last exon was cut off by a stop codon, do HSP adjustments: from the stop codon 
site, check the HSP at its left side and HSP at its right side, and remove the HSP that 
has lower identity. Then re-compute the intron regions according to the new set of HSPs 
and re-compute introns for each intron regions and so on.
- further modifications:
* For last exon, if there's a second-to-last exon, try extend that exon to its next stop.
Compute the global alignment between this newly-deduced gene (with the original second-to-
last exon as the last exon) and the query. Compare its PID with the global alignment 
between the gene in previous method (with repair etc.) and the query. Choose the one with 
higher PID as the final prediction.
- A bug fix:
Input function that reads the input target DNA database was fixed, so that now very long 
lines in the input file can be read correctly.

V112:
- fixed a bug for "newHSP" to extend beyond its left/right limit during repairing mid exons 
by adjusting the limits to be in frame
- for intron regions between two HSPs that have in-frame stop codon in between, also check 
the "no-intron" case, if its PID is the highest, then do the removeHSP stuff immediately 
before trying to recompute intron regions and exons
- if the initial prediction contains a single exon, treat it the same as multi-exon case, 
with RepairHeadExon and RepairTailExon one at a time (not trying to repair at both ends at 
the same time)
- reduce alternatives: 
(1) change GlobalAlignment to maximize score instead of PID; 
(2) change the alignment ranking measure to first score then PID (instead of PID alone);
All comparisions between alignments are now done by comparing score then PID.
(3) if local HSP-based spliced sequence alignments result to multiple same highest ranks 
(alternatives with same score and PID), then check optimal alignment betweent the spliced 
sequence and corresponding query segment, select the one with highest score then PID.
* NOTE the global PID computation is changed, so it now seeks the alignment that maximizes 
score and the resulting PID may be slightly lower.

v113:
- "reduce alternatives" modified (change the alignment comparison again):
(1) alignment aims to maximize PID
(2) compare alignments by PID first, then BLOSUM62 score
- added additional line of output in raw output
"Gene:ID=gff_gene_id|gene start-gene end|strand|rank|score|PID"

v114:
- cosmetic fix for the final "Gene:ID" line in raw output

v115:
- fixed a bug in computing optimal alignment in each intron region (when there're more than 
one alternative donor-acceptor pairs in an intron region that have the same top rank), to 
consider the case of "no-intron"
- updated help function to specify ungapped setting as default

v116:
- bug fix, actually use PID before alignment_score to compare alignments (misplaced header)

v117:
- update/clean up output

v118:
- update output, add "NEW EXON..." lines to report novel exons not based on BLAST HSPs

v119:
- bug fix, adjust len in GetSubstrFromVecStrs_*_ForRepair for short sequence

v120:
- output additional info on 3' exon repair (alignment and PID for each of the 3 options: 
(1)initial end exon; (2) extended end exon; (3) second-to-last exon as end)

v121:
- in genBlastG, for each query gene, load one chromosome at a time and process all groups 
on that chromosome only, then move on to the next chromosome (to be able to handle larger 
genomes that can't fit into memory), assume memory is large enough to comfortably hold 
any single chromosome (largest human chromosome is about 250M)

v122:
- update genBlast to include parsing for both wuBlast and Blast, enabling the "-P" option
- update output PID and coverage numbers to up to 2 digits after decimal point

v123:
- modify parsing of wublast/blast output to get proper database sequence name from ">" line
- add sequence name checking to verify sequence input is properly done, instead of giving 
exception later

v124:
- fix bug in finding(extending) alignment during repair process, for small contigs where 
there is not enough base pair left in sequence, to ensure in-frame translation
- minor correction on output format, to get proper final_start_site/final_end_site when 
gff option is not enabled

v125:
- change DNA sequence loading to loading all DNA regions (for all groups in the same gene) 
at once, in order to speed up some

v126:
- add the file position index data structure to speed up sequence scanning process

v127:
- a bug fix, to compute in-frame extension length on DNA sequence for head/tail exon repair

v128:
- a bug fix, make sure there's enough base pairs left on DNA sequence when examining 
acceptors/donors for possible internal exons, in "PostProcessSpliceSites"

v129:
- add "-pid" option, for global alignment PID computation (global alignment between query 
and prediction), now, by default it is off.

v130:
- add "-scodon" option, to allow searching for start codon inside HSP region (from first 
HSP, search downstream) up to certain number of base pairs, default is 15.
- change gene end site to include the stop codon at the end.

v131:
- bug fix, update the utility function HasInFrameStop to avoid failing last exon due to 
included stop codon at the end.

v132:
- bug fix, check fgets return value when reading chromosomes to ensure something is read

v133:
- fixed a PID-reporting format issue in gff output

v134:
- In the merging process, remove HSPs that are properly contained in other HSPs 
(resulted from previous rounds of merging)

v135:
- minor bug fix in gff and cDNA output format

v136:
- bug fix, in repair mode, after removing HSPs, make sure start_site does not come after end_site
- reset "-P" option to use "blast" as the default search program

v137:
- silence g++ compile warnings by explicit type cast and replace string::npos with -1
- add macro for Mac portability (for large file operations)

v138:
- add "-W" option to allow user setting on Blast/WuBlast -W option (set word size).
- clean up gBlast.cpp for esthetics reason

v139
- Changed all 'endl' to "\n" to prevent short writes.