Fig. 2. Combination of multiple absolute quantitation modules for CNV detection.

a Stochastic error in copy number quantitation was reduced by increasing the number of quantitation modules in the gene. ERBB2 ploidy was calculated as 2 times the ratio between the mean of UMI family counts from modules in ERBB2 and in the reference. Because there are multiple possibilities of module down-selection, here each datapoint represents the average of 30 randomized selections. 123 modules served as the reference for ploidy calculation. The CV of 1 module was lower than that in Fig. 1d, because only 1 module was used for both the reference and ERBB2. b Discriminating 2.05 and 2.00 ploidy ERBB2 samples using QASeq. The P-value calculated using the two-sided t-test was 0.0013. 2.00 ploidy sample was a normal PBMC DNA sample (ZB4) tested in four independent experiments; 2.05 ploidy sample was prepared by mixing the same normal PBMC DNA sample (ZB4) and ERBB2-positive cell-line (SK-BR-3) DNA, and was tested in two independent experiments. c Concordance of QASeq ERBB2 ploidy with IHC. For the sample with discordance between QASeq and IHC, ddPCR agreed with QASeq; this might be a result of ERBB2 over-experession at protein level due to changes in promoters or enhancers. d Concordance of QASeq ERBB2 ploidy with ddPCR. For the three samples with discordance between QASeq and ddPCR, IHC agreed with QASeq. e Ploidy of each QASeq quantitation module for a healthy PBMC donor (top) and for a breast cancer fresh/frozen (FF) tumor section (bottom). Genes with called CNVs are displayed in orange, genes within expected variation are shown in black. The genes in gray have <3 modules per gene in QASeq, and thus cannot be used to calculate gene ploidy using Mann–Whitney U test. f QASeq CNV calls for 18 fresh/frozen breast tumor sections. Two non-adjacent tumor sections were tested from patients p15 and p16. Between 5 and 8 ng DNA input was used based on availability.