Does anyone know a recent comparison done of the BWA-mem and Berkeley SNAP aligner?
Especially if it leads to similar True Postive (TP), True Negative (TN) variant call ratios for snp, indel, complex and structural variants. (by variant calling the BAM files produced by BWA and SNAP using GATK/ Freebayes and structural variant callers).
Also it would be nice if the actual speed was compared between BWA-mem and SNAP in the comparison.
The authors of SNAP claim "that it is 3-20x faster and just as accurate as existing tools like BWA-mem".
See also this presentation: http://snap.cs.berkeley.edu/downloads//ASHG%20SNAP%20Pandya%2010-19-2014.pptx
SNAP would be a good alternative to bwa-mem if someone could independently confirm these precision, recall and speed numbers, for human and maybe a second non human genome.
If you care about >15bp longer indels or break points, SNAP is not for you; if you don't, SNAP is pretty competitive in terms of mapping accuracy. Don't know how it plays with variant callers.
I would ask Brad Chapman about this. He's done so many comparisons with variant callers and settings that I imagine comparing aligners would be simple for him.
BTW, I wouldn't be surprised if the SNAP claims are correct since I recall it requiring vastly more resources than bwa mem.