To evaluate the power of our computational approach in cancer genomics studies, we used it to analyze six metastatic melanoma cell lines derived from metastasis tumor biopsies of stage IV melanoma patients and six blood samples from healthy donors used as normal controls.

Whole-exome sequencing (WES) represents an effective experimental strategy for the capture of the genetic variation falling in the coding portion of the genome. So far, WES has been successfully used for the detection of single nucleotide variants, small insertions/deletions, and breakpoints of structural variation. However, WES is refractory to the discovery of copy number variants (CNVs): the non- uniform read-depth among the captured regions and the sparse nature of the target make WES data unsuitable for traditional CNV detection algorithms. To overcome these limits in the discovery of CNVs we studied the statistical properties and biases of depth of coverage of captured sequencing data and we developed a three-step normalization procedure that mitigates the non-uniform read-depth among the captured regions. Moreover, we developed a novel Hidden Markov Model based algorithm that, by exploiting the sparseness of the target, is able to detect large alterations as well as small and highly isolated altered regions, outperforming the state of the art segmentation algorithm. We combined the normalization and segmentation algorithms with a calling method that classifies each genomic region into copy number states and we packaged them in a novel software tool that we named EXCAVATOR (EXome Copy number Alterations/Variations annotATOR). To demonstrate the power and versatility of our tool we used EXCAVATOR for the analysis of three WES datasets (a population dataset, a cancer dataset and a intellectual disability dataset) and we compared the results with corresponding copy number profiles generated by Single Nucleotide Polymorphism array technology. The results we obtain show that EXCAVATOR can be a valuable tool for the investigation of CNVs in large-scale ongoing projects (such as the 1000 Genomes Project and the Cancer Genome Project) as well as in routine research and diagnostic activities.