run_whole_alignment doesn't output chromosomal BAMs

[cmarkello]

@glennhickey
The run_whole_alignment function in vg_map.py with bam_output=True or surject=True doesn't appear to process and return merged bams by contig, but instead just merges all bam chunks from all contigs together.

https://github.com/vgteam/toil-vg/blob/master/src/toil_vg/vg_map.py#L401

[glennhickey]

Yup, doesn't look like that logic's implemented for BAM. Surprised it doesn't crash looking at the code. It'd need the BAM equivalent of this function https://github.com/vgteam/toil-vg/blob/master/src/toil_vg/vg_map.py#L653-L655 to be run.

[cmarkello]

Wouldn't the following also need to be refactored in order to take into account the chunked bam output of run_chunk_alignment?

https://github.com/vgteam/toil-vg/blob/master/src/toil_vg/vg_map.py#L456-L457
https://github.com/vgteam/toil-vg/blob/master/src/toil_vg/vg_map.py#L463-L464

[glennhickey]

I imagine so. Shuffling the GAM chunks by chromosome was a necessity for the old vg call pipeline, not so much anymore. A whole genome GAM can now be split into (unsorted) chromsome chunks with vg chunk -M in a few hours (I think). You can extract a chromosome out of a whole-genome BAM even faster with samtools view. Though, using samtools at least, you'd probably need to index the BAM to efficiently do many chromosomes at once.

[cmarkello]

Yeah I've been working on a solution that did chromosome splitting per chunk. But I agree, I think splitting by contig after merging the BAM would be more efficient and simpler than merging chunked contig BAMs together.