Question about allignment of reads to duplicate regions

[watersss]

Hello,

Perhaps you can help me out with an issue I am having. In my current project I am trying to identify a mutated gene containing a deletion. The mutated genome was sequenced using nextgen paired end sequencing and the reads were aligned using Bowtie. I have identified a candidate gene, however, the reference genome shows that the candidate gene has a duplicate within 200 kilobases.

My question concerns to what degree this duplicate could be affecting the alignment. Is it likely that the observed coverage gap in the candidate is due only to the effect that the duplicate might have on the overall alignment?

Thanks

[dpryan]

Why not just simulate some reads covering that area with and without the deletion, align them, and see if the duplicate is screwing with their mapability?

[dcameron]

An exact duplicate with default alignment settings (return only the best hit, and randomly choose one if there are multiple equally good hits) would result in reads aligning to either copy of candidate being randomly assigned. Ignoring all other factors (contamination, sequencing errors, sequences elsewhere in the genome, insufficient coverage, etc), you would expect the aligned reads to show a deletion signal in both your candidate and the duplicate if they have the same sequence around the deletion region.

[andylemire]

How big is the deletion, relative to the library size? Are we talking indel-sized, or deficiency-sized? dpryan's suggestion works for deletions contained within a read, but if it's a large chunk, larger than the library, then you'd be simulating coverage differences, which would give you whatever answer you want.

I agree with dcameron that you should see the decrease in coverage around the deletion in both loci with the random assignment of perfect alignments to both, assuming the "duplicate" is a perfect sequence match. However, I've seen bowtie2 alignments (SE, not PE) with a strong preference towards the leftmost occurrence of a duplicated sequence in a reference genome, suggesting it wasn't really random assignment.

Perhaps local SNPs near flanking regions at either loci could help resolve whether the alignments were placed correctly or not.

[dpryan]

How big is the deletion, relative to the library size? Are we talking indel-sized, or deficiency-sized? dpryan's suggestion works for deletions contained within a read, but if it's a large chunk, larger than the library, then you'd be simulating coverage differences, which would give you whatever answer you want.

Good point! I should have mentioned that I was assuming a smallish deletion.