partition.pl vs partition_gmap.pl vs partition_gmap.py

[diriano]

Dear @tangerzhang and @tanghaibao,

I see some inconsistencies between the files partition.pl vs partition_gmap.pl vs partition_gmap.py. I would like to know whether this is intended or whether it is an error.

In partition_gmap.pl (d5bb1e5) line 65 reads

next if(exists($rdb{$rname})); ### only retain single-end reads

The comment is not appropriate, as this will not keep single-end reads, I will only keep the first alignment of a given read.
The same line is present in partition.pl (b49ddea) line 51.

Nothing like that is present in partition_gmap.py (80ce6a9). This, I understand, would be the correct behavior, keeping all alignments from contigs originating from the same chromosome.

I have been using partition.pl with a sugarcane genome of 11Gb and a pruned bam of 250GB, this ran out of memory, on a machine with 0.5TB RAM. I have rewritten partition.pl to use a lot less memory, at least 20x less than your version. This version uses bioperl to index the assembled genome, and goes a single time through the streamed BAM file (without loading it in memory), available here. And another version that also streams the BAM file, both only for a set of contigs in a given chromosome

I will appreciate your comments on this.
Thanks a lot in advance.
Best,
Diego

[diriano]

Another point. in partition.pl, line 76

next if($count>1);

Only allow to assign a contig to one and only one chromosome from the reference. Each time the script is run, a different chromosome could be chosen. Perhaps a better way to chose would be to the one with the highest number of hits, something similar to what is done in partition_gmap.py in lines 53 - 57.
Any thoughts?
Thanks