Fig. 3.

Graphical representation of structural variation detection algorithm. Initially, Aldy compares the copy number neutral input coverage with the reference sample coverage in order to establish proper depth of coverage. Only one reference sample is needed for each sequencing technology, and Aldy comes with such data by default. In the second step, the coverage of the sample is normalized (this is equivalent to establishing the function cn_s). The first two steps are described in Supplementary Note 1. In the third step, we segment the gene into the regions, and estimate the average copy number for each region (i.e., establish the vector cn). Finally, we infer the set of structural configurations (i.e., vectors v) which best “fit” the observed segmented coverage cn via integer programming