Bug #12245
open
pasta rotini-fastj does not handle tagset without ending newline
Description
There is a bug in the pasta rotini-fastj conversion where if the tagset provided does not have a trailing newline, the last tag will be ignored.
For example, the following should work but doesn't:
pasta -action rotini-fastj \ -start 0 \ -tilepath 0000 \ -chrom chr1 \ -build hg19 \ -i stage/hu034DB1-GS00253-DNA_A02/0000.pa \ -assembly <( l7g assembly assembly.00.hg19.fw.gz 0000 ) \ -tag <( samtools faidx tagset.fa.gz 0000.00 | egrep -v '^>' | tr -d '\n' | fold -w 24 ) > stage/hu034DB1-GS00253-DNA_A02/0000.fj
The culprit being the
samtools faidx tagset.fa.gz 0000.00 | egrep -v '^>' | tr -d '\n' | fold -w 24.
As a workaround, adding an extra newline will correct the issue:
pasta -action rotini-fastj \ -start 0 \ -tilepath 0000 \ -chrom chr1 \ -build hg19 \ -i stage/hu034DB1-GS00253-DNA_A02/0000.pa \ -assembly <( l7g assembly assembly.00.hg19.fw.gz 0000 ) \ -tag <( cat <( samtools faidx tagset.fa.gz 0000.00 | egrep -v '^>' | tr -d '\n' | fold -w 24 ) <( echo "" ) ) > stage/hu034DB1-GS00253-DNA_A02/0000.fj
The place to look is the readTag function in pasta_fastj.go and where the g.TagFinished flag is referenced but the details of what's wrong need to be investigated.
This bug specifically happens when there is a variant on the next to last tag in the stream being converted, as is the case for data set hu034DB1-GS00253-DNA_A02 in tilepath 0000.
Updated by Abram Connelly over 6 years ago
Note this happens with pasta version 0.2.6 and lower.
Updated by Sarah Zaranek over 5 years ago
Currently the cwl pipeline using 0.2.6. Need to insure that this fix in all relevant code.
Updated by Jiayong Li over 5 years ago
In convert2fastj of gvcf src code, there's the line
$pasta -action rotini-fastj -start $tilepath_start0 -tilepath $tilepath -chrom $chrom -build $ref \
-i $odir/$tilepath.pa \
-assembly <( $tile_assembly tilepath $afn $tilepath ) \
-tag <( cat <( samtools faidx $tagdir $tilepath.00 | egrep -v '^>' | tr -d '\n' | fold -w 24 ) <( echo ) ) > $odir/$tilepath.fj
We need to verify 'echo' is functionally the same as 'echo ""'.
Updated by Anonymous over 5 years ago
Pasta 0.2.8 can be found in the l7g abramVersion branch. This needs to be tested thoroughly.