BWA and masking the reference genome

[jullee]

I am using BWA to align paired end Illumina reads to a published reference (where each of our target species' chromosomes is represented as a large scaffold). However, there appears to be a large colinear stretch representing about 70% of the chloroplast in one of these scaffolds. As recovering the chloroplast is a major goal of my analysis, I am thinking about masking this region of the scaffold in question (when I do the alignment and don't mask, I get very low coverage of the cp after filtering out ambiguous reads).

I know I can mask the reference using bedtools and I think in this case I want to use hard masking (replace with Ns) but I am wondering if this will achieve the desired result: e.g. BWA will not allow any reads to map to this region. Can anyone clarify how BWA treats masked regions of the reference?

Thanks!

[dpryan]

Yes, if you hard mask then you'll get no alignments to that region. That seems like a reasonable thing to do given your goals, though keep in mind that if there are stretches of genomic DNA that have relatively high homology to the chloroplast sequence then you'll get a somewhat increased false-positive alignment rate. The alternative is to not mask and then just extract the ambiguous alignments to that region and then convert them back to fastq (presumably for assembly).

[jullee]

Thanks dpryan! That answers my question.