We sequenced 71.1 kb of cancer mutational hotspots in DNA samples from a primary colon adenocarcinoma with its matching xenograft, a breast primary carcinoma with its matching xenograft, an ovarian carcinoma xenograft, a sarcoma xenograft (Materials and methods) and the matching germline DNA derived from all four patients' blood. The xenograft samples had been passaged between two and seven times in immunodefficient mice (Materials and methods). To call the significant mutations, we trained the method with a calibration sample sequenced in the same run as the clinical sample (Table S7 in Additional file 2). As shown above, the performance of UDT-Seq is significantly better for higher prevalence mutations. For this reason, we restricted our analysis to the mutations identified at a prevalence of 5% or greater. The somatic mutations were then called by analyzing the differences between cancer and germline DNA mutations. We considered all variants in a cancer sample that passed statistical assessment (Materials and methods) as potential somatic mutations. We then retained the mutations for which the corresponding position in the germline sample was covered (> 10 reads)
