Aligner for SNP calling

[honey]

I have a data which have been sequenced on Illumina Hi-seq 2000. I want to study sequence variation between two samples. We have been generally using BWA, Bowtie to align to human genome. Can I use BWA for this purpose if so is there any specific adjustments in running parameters of BWA to edit if my aim is to find variations in sequences? In general which will be the best aligner for seq variation detection calls.

[sdriscoll]

BWA or Bowtie2. Both of those can do gapped alignments which is what you want as a starting point for SNP calling. BWA does this by default. I believe it's also wise to only allow unique alignments. Otherwise I run BWA with default settings aside from controlling insert size on PE data. Maybe someone else will have more detailed suggestions.

[lh3]

Bowtie2 is great, but if you use samtools for indel calling, remember to run an indel realigner or left-aligner; otherwise, the samtools indel caller won't work. I don't know about other indel callers.

[lukas1848]

I used stampy&bwa for my mapping. Is there any reason to choose bowtie2 over stampy?

[Heisman]

Bowtie2 is great, but if you use samtools for indel calling, remember to run an indel realigner or left-aligner; otherwise, the samtools indel caller won't work. I don't know about other indel callers.

Pretty sure Dindel by default does left-aligning. Not to hijack this thread but I know you made a post somewhere recently saying it was somewhat up in the air to you regarding SAMtools vs. GATK vs. Dindel for indel calling. If you see this I'm curious if you think that's true for whole genome and targeted capture data or if there is a distinction between them?

[adaptivegenome]

I used stampy&bwa for my mapping. Is there any reason to choose bowtie2 over stampy?

stampy is slower but better for detection of large indels (>10bp), otherwise BOWTIE2 is pretty awesome

[adaptivegenome]

Pretty sure Dindel by default does left-aligning. Not to hijack this thread but I know you made a post somewhere recently saying it was somewhat up in the air to you regarding SAMtools vs. GATK vs. Dindel for indel calling. If you see this I'm curious if you think that's true for whole genome and targeted capture data or if there is a distinction between them?

We have tried GATK versus DINDEL and we do not see much difference. I am guess that GATK's method is based on DINDEL even though there is not an option anymore for the DINDEL algorithm.

[lh3]

GATK never implements the exact dindel model. It just borrowed the fundamental idea and the basic model, but many implementation details are very different.

I think for indel calling there is a big room for further improvements. Most numbers in publications and those I have heard are inconsistent with each other. Everyone (including me) is arguing their methods are the best, but I do not believe that before the community reach a consensus.