Figure 4

Discriminative frequent CNV patterns of Hapmap data. The plot shows leave-one-out classification accuracy on a portion of the 3 populations using frequent CNV patterns selected by remaining individuals. (a) Leave-one-out accuracy on 90% of the samples using patterns selected by 10% of the samples. (b) Leave-one-out accuracy on one third of the samples using patterns selected by two thirds of the samples. The x-axis shows the number of patterns used for classification where the patterns are first ranked based on their population specificity. SubPatCNV-L stands for SubPatCNV patterns with size at least 2; JISTIC-L and JISTIC-S stands for JISTIC patterns with size at least 2 or all patterns including singletons respectively.