bug: bbmap aligned reads have chr "*" when outputting in BAM format

[eboyden]

Using bbtools v40.01

bbmap output in standard SAM format (out=out.sam) works just fine, but running the same command except with BAM output (out=out.bam) produces identical alignments except in every case field 3 (chr) is *. The other fields and tags, including alignment coordinates, CIGARs, map quality scores, and bitflags are identical.

[btpsj]

Can you give the actual command or parameters you ran bbmap.sh with so its easier to replicate your issue?

[eboyden]

bbmap.sh -Xmx8g -eoom usemodulo int fast maxindel=100 minid=0.85 ambig=toss kbp po pairlen=1000 apd=-25 deterministic ordered trd sam=1.3 rgid=$r rgLB=$s rgPL=ILLUMINA rgPU=$r rgSM=$s path=$index_bbmap in=stdin.fq out=out.bam 2> bbmap.log

[btpsj]

Thank you, I'm able to reproduce the observed behavior (actually a couple different ways) but to make sure its the right one, can you confirm whether or not if the reference fasta you are using has header lines with description text? For example

>chr1 description
CACACACACCACA

if after the chromosome name there is a space and some amount of description text following it in your reference file you used to build the index.

[eboyden]

Yes that is the case:

>chr1  AC:CM000663.2  gi:568336023  LN:248956422  rl:Chromosome  M5:6aef897c3d6ff0c78aff06ac189178dd  AS:GRCh38
NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN

[btpsj]

Great, so that is useful in figuring out the source of the problem, hopefully the only one. While a fix is underway, a non-obvious way to deal with with is using the nativebamout=f parameter on bbmap.sh. This isn't documented in the help text, but if you have samtools in your $PATH it will use that to do the conversion. I've used a minimized example that shows that it does fix that problem. Otherwise, if a commit comes in that fixes the issue I'll message here.

[btpsj]

The latest commit (df45e18) should have a fix.