In order to estimate the error rate, train the statistical model and measure the performance of the assay, we prepared calibration samples by pooling four Coriell DNA samples (NA12156, NA12878, NA18507, or NA19240). These Coriell samples have previously been subjected to exome sequencing and thus the positions of coding polymorphisms are known [15]. We pooled these samples four times, permuting the relative concentration of the samples (1%, 5%, 20% and 74%), to obtain four different calibration samples referred to as CAL-A to CAL-D. We selected 40 calibration coding SNPs based on the fact that they were homozygous for the alternate allele in one of the four samples and homozygous for the reference allele in the three others. We repeated this selection four times, permuting the four samples, resulting in a total of 160 calibration SNPs. We successfully designed primer pairs to amplify 200-bp fragments around 159 of these SNPs. Because we perform direct sequencing from the amplicon ends (Figure S1b in Additional file 1), we made sure that the calibration SNPs were located at various positions throughout the 200-bp amplicons.
