Best aligner for both long and short read length

[Greg_owens]

Hi,

I'm trying to compare sequences from several different libraries sequenced at different times. Consequently they differ in read length from about 35bp to 150bp. I'm planning on using BWA but I'm sure whether I should use BWA mem or BWA backtrack. BWA mem seems superior for reads 70bp+ but some of my samples are shorter than that. I'd also like to use the same aligner for all my samples for consistency sake.

So the question is how does BWA mem do on short read samples? The manual says it is not recommended, but how much worse is it? Is there another good option for aligning libraries with such different read sizes?

Thanks

[Brian Bushnell]

BBMap does well with both long and short reads, though longer reads will always allow higher confidence because they contain more information. It was originally designed to handle high-error-rate 25bp reads for the SOLiD 4 platform, and now can handle PacBio reads (also high error rate) up to 6kbp, so it's pretty flexible in terms of length.

[luc]

BBMap works great for me - I like it for all the analytic options, the speed, and the HQ results. I have never attempted to align anything shorter than 50 bases with it though.
The goal of consistency of the tools seems to be misplaced for this purpose. I would feel more comfortable using the optimal tools for each of the different data sets.

[westerman]

@luc: I disagree about 'The goal of consistency of the tools seems to be misplaced'. If I am doing comparative studies between different libraries (which is what Greg_owens appears to be doing) then I want to use the same tool (and version of the tool) for all of my studies. Otherwise I can not be certain if differences in the comparisons are due to library differences (what I want to find out) or tool differences.

That said, I think that we would both agree in recommending BBMap to Greg_owens as the one tool to use for his comparative studies.