Aligment Of Reads To A Bisulfite Converted Genome

Question

[Tonig]

Hi everybody.

I got 350 samples from a methylation-seq experiment, these samples were designed using BLAT in order to look form single-matching regions. My question is, when I use bisulfite aligment programs like RMAP (RMAPBS command) or BSMAP (that is, I try to align my reads agains a bisulfite converted genome), I get a different number of reads that every one of them align exactly to an unique position of this converted genome, but is a different number of reads compared with those that I get aligning against the non bisulfite converted genome using BLAT. Is this normal? I mean, the same number of reads that align in unique positions of the genome, don't align in the same number in the bisulfite converted genome. It is possible? why?

Thanks in advance

[Alastair Kerr]

My best guess would be the effects of mismatches and (if diploid) allelic variants. As bisulphite may not convert 100% of the methlyl-C to U, mismatches are possible. I have not used RMAP or BSMAP: how do they handle mismatches? Are there mismatches in the extra reads that map?

[brentp]

I get a different number of reads that every one of them align exactly to an unique position of this converted genome, but is a different number of reads compared with those that I get aligning against the non bisulfite converted genome using BLAT.

could you clarify what you mean? if you're saying that you have a different number of reads mapped by BLAT (to the normal reference) and by an aligner designed to map BS-Seq reads, that doesn't seem surprising at all.

[Tonig]

Thanks Alastair, I'll try to contact the developers of both programs but maybe this is one option

Answer 1

[Charles Warden]

Different aligners should typically provide slightly different results (if they use identical algorithms, then you can't really justify publishing a new algorithm under a different name).

For example, my understanding is that Bismark is more conservative than BSMAP, so I wouldn't expect the same number of reads (or, potentially more importantly, the percentage methylation at each site).

A normal BLAT search should definitely be different. In general, BS-Seq aligners should actually be using at least 2 reference sequences at the same time (a bisulfite converted one and a non-converted one). A BLAT search should only work on reads corresponding to regions that weren't altered during the bisulfite conversion. Otherwise, you could just use any short-read aligner (BWA, Bowtie, etc.).

[Charles Warden]

Might also be worth checking out this article:

http://www.ncbi.nlm.nih.gov/pubmed/22344695