SNP calling problems

[jomaco]

Hi

I am carrying out alignments using BWA and Bowtie and then using samtools to produce some fairly large SAM files (>100G) followed by calling SNPs and producing VCF files.

I have found that BWA seems to be taking many days longer to run than Bowtie. However, for test sequences BWA seemed to be sufficient for our purposes, the read depth of the aligned sequences to the reference sequences was sufficient to call SNPs which pass our statistical tests. Bowtie, however, was much faster but results often only had a read depth of 5 or so and much fewer SNPs were called and those that were did not pass stats tests.

I thought reducing the stringeny of Bowtie somehow might help; I used -v 3 and -y with resulted in a read depth which seemed to be slightly higher. A few more SNPs were called but this was not sufficient.

Are there any other tools I can use which will be faster than BWA or other Bowtie parameters I could explore? (I am expecting a read depth more towards 50-60) as was originally possible for the test sequences when using BWA.

For certain datasets am I also receiving errors in the sai -> sam conversion stage. With BWA taking so long to run and then producing these errors it is taking much longer to carry out these tests and then re-run them than originally expected giving our success with the test sequences. There are certain datasets where I just cannot get past this stage. Now the input files contain multiple sequences instead of just one, could this be an issue? Would perhaps piping the output of BWA to samtoools for the production of the SAM file help to stop this error?

I am really at loss as to why scaling up the process from the test sequences has resulted in these problems. I can give more biological context if required. Hope you can help!

Thanks

Jom