graph for each sample and for each algorithm. Using the cnvPartition algorithm, we detected 16, 10, 7, 13 and 11 CNV calls for the five samples with increasing DNA quantity. With the PennCNV algorithm without wave adjustment, we detected 24, 14, 14, 11 and 13 CNV calls for the five samples. Clearly, for both algorithms, the sample with the lowest DNA quantity (and the strongest wavy pattern) had more CNV calls, reflecting possible false positive calls. The concordance between the five samples was generally poor (Supplementary Tables 2 and 3).
