SNP-calling was highly accurate for SNPs with a frequency > 3% for all sequencing platforms simulated. Roche/454 showed a slightly lower true positive rate (i.e. an increased false negative rate) than the Illumina simulations. When using VarScan with a M.A.Q. of 20, 379 out of 400 SNPs were identified (86, 97, 97, and 97 for frequencies 1, 3, 4 and 7%, respectively). In contrast, all 400 SNPs were identified from both Illumina simulations. Upon closer inspection, 11 of the false negatives with true frequency of 1% failed to be supported by five reads. All the remaining false negatives were associated with homopolymer indel errors. Inspection of the pileup file showed these SNPs were contained in the data; however VarScan reported them as indels instead.
