samtools mpileup & metagenomic data

[jmartin]

I would like to call snps on metagenomic collections of haploid genomes (bacteria), and I've tried samtools mpileup (+ vcfutils varFilter) just to see how it would handle it. What I see is that mpileup assigns scores to the 3 possible diploid genotypes at each loci, and I'm trying to figure out exactly what its doing.

The reads covering each loci are coming from a population of haploid individuals, but I suspect that mpileup is making calls into the 3 diploid genotypes based purely on allele frequency. If it sees the reference allele & the next most common allele in about a 50/50 mix, it calls it heterozygous (REF/ALT). And then I figured it probably would accept some small range of frequencies around 50/50 (maybe from 45% to 55% mix) where it would still consider that REF/ALT, and then every other frequency combination it would call either homozygous REF or ALT (REF/REF or ALT/ALT).

This is clearly not ideal for my situation, but I was thinking it would probably be able show me common variants in my population.

But this is all guesswork on my part, I was hoping someone on here might be able to explain a little better just what mpileup is doing with this metagenomic data.

[Torst]

FreeBayes may be more flexible for your needs (mixture of haploid genomes):

http://bioinformatics.bc.edu/marthlab/FreeBayes

We are looking at using it for our bacterial SNP studies, as mpileup and GATK all make assumptions about diploid.