Hi all,

I have a project in which I am attempting to generate a library of deletions in a gene; the deletions vary in size from single codons up to about 3 kilobases, with a total of about 1 million variants. I'm planning to do some screens with this library, sequence the start/end populations via NGS, and map the reads against a reference sequence so as to determine which deletions are present in each population.

I'm quite new to bioinformatics in general, so I was hoping I could get some advice on how best to approach the last bit. I have used Bowtie2 in the past for other library-based projects, and my understanding is that it is well-suited for detecting deletions. However, in my limited experience using it for this project, I am finding that it seems to accurately map small deletions (tens of nucleotides / shorter than the length of an NGS read), but seems to stumble when handed large deletions (longer than the length of an NGS read / several kilobases). In these latter cases, the SAM output file contains sequences with CIGAR strings indicating a series of mismatches and short deletions (for example, a read containing a junction corresponding to a deletion of 2700 bps has a CIGAR tag of 4M1D3M6D1M1D1M2D1M1D1M2D1M3I3M1D1M1D126M). I have tried to address this problem by lowering the gap penalty scores to 1, which results in a different CIGAR string but basically the same outcome.

As I said before I'm very inexperienced at this, and I wouldn't be surprised if there's a simple fix that I'm unaware of (please feel free to reply if you happen to know!). However, supposing that this does in fact reflect a fundamental limitation of Bowtie2, does anyone know of an alignment algorithm that would be ideal for the purpose described?