SNPest: a probabilistic graphical model for estimating genotypes

Table 1 Performance of SNPest, GeMS and FreeBayes on real data

SNPest and GeMS use various fractions of the available data (from maximum 5 reads per site to maximum 60 (in this case all) reads per position), and FreeBayes is only run using all available data. The REL606 strain of E. coli was sequenced on the MiSeq platform to an average depth of 27X. Residual adapters were removed using AdapterRemoval, and the cleaned reads were mapped using two different mappers, Bowtie2 and BWA-PSSM. No SNPs are expected in this mapping, as we are mapping a known sequence back to itself. SNPest used the reference genome as a prior (see Additional file 1 for more results). All: All SNP candidates. QC: Number of SNPs after filtering on quality (SNPest and FreeBayes: Genotype quality of >30. GeMS: Dixon Q-test p-value <0.01).