The role of chromosomal rearrangements in cancer is well established, and some recurrent patterns are known. The mode of action of a number of effective anti-cancer drugs (e.g., imatinib, crizotinib) is to inhibit the products of gene fusions resulting from such chromosomal rearrangements. Recent studies suggest that common rearrangements in tumor genomes may be more prevalent than previously thought. Thus, elucidation of recurrent rearrangements, and acquisition of systematic knowledge and analysis of such rearrangements is a promising strategy that can further our long term goal of identifying novel targets and corresponding therapeutic opportunities. However, the lack of reliable methods for detecting cancer-related chromosomal rearrangements (often called structural variants) represents a significant stumbling block to progress toward this goal. The problems with current methods are centered around their inability to integrate different types of evidence, and their lack of comprehensive handling of sequencing errors and biases. The objective of this proposal is to develop software that overcomes these problems related to the identification of structural and other variants in tumor genomes. We will develop a novel algorithmic framework and functional software to improve predictions of cancer variants including copy number and breakpoint resolution by filtering genome biases and integrating all available sequencing evidence. Our tools will report genome changes of all types, from structural to single nucleotide variants, in a single package. This will have an important added benefit of significantly reducing the time of the overall data analysis of tumor genomes. Comparing the breakpoints and other mutations between the tumor and normal genome will provide information regarding common and tumor- specific genomic patterns, indicating possible factors associated with cancer pre-disposition and somatic changes driving tumorigenesis. We will validate our approach utilizing available experimental data, and we will distribute the software using an open source model. This project will support active participation of graduate and undergraduate students, and their involvement is likely to generate interest and motivation toward careers related to a biomedical field. The proposed research will build on the successful early tests of our integrated approach to structural variant prediction. We anticipate that this work will result in the generation of computational tools that will permit robust identification of genome variants in cancer cells, which will be key to understanding tumorigenesis and to identifying targets for intervention in an individual tumor-specific manner.
The role of chromosomal rearrangements in cancer is well established but current algorithms predicting rearrangements from tumor sequencing data tend to generate divergent sets of predictions with high error rates. We will combine several methods to produce a computational tool to reliably identify rearrangements and other changes in tumor genomes. Such tool will help cancer researchers to determine commonalities, differences, and emergent mutational patterns across diverse cancer types, which may lead to finding novel targets and treatment options.
