Fig. 1

RC data distribution, biases normalization and CNV prediction capabilities. Panel a shows the histograms of RC data from Illumina Platinum WGS experiments for different sequencing coverages with the superimposed Poisson (dashed lines) and negative binomial (solid lines) distributions. The distance between RC and Poisson (DP) and negative binomial (DNB) distributions are calculated with the Kolmogorov-Smirnov statistic D. Panels b and d report the correlation between RC data and GC content % and mappability. Panels c and e show the effect of median normalization approach on GC content % and mappability. Panels f, g and h summarize the capability of RC data to detect deletions (f), duplications (g) and false positive events (h) as a function of sequencing coverage and window size. The TPR of panels f and g are the proportion of correctly predicted simulated deletions and duplications respectively. The FPR of panel h is the proportion of 2 copies regions predicted as deletions or duplications. Panel i show the association (regression coefficient) between real and predicted number of copies as a function of sequencing coverage and window size