Algorithmic approaches to systems biology, data integration, and evolution

National Library Of Medicine

Investigators

Linked publications, trials & patents

Abstract

My group continued to develop and apply computational methods that utilize and integrate large data sets with a focus on gene regulation and diseases. Cancer genomes accumulate a large number of somatic mutations resulting from various endogenous and exogenous causes, including spontaneous deamination of cytosines, mutations triggered by carcinogenic exposures, or cancer-related aberrations of the DNA maintenance machinery. These mutations are mostly harmless passenger mutations, but analyses of their patterns can provide useful information regarding mutational processes acting on cancer genomes. Indeed, different mutagenic processes often leave characteristic mutation imprints in cancer genomes. Identifying processes shaping the mutational landscape of cancer is an important step towards understanding tumorgenesis and has the potential to inform therapeutic and preventative measures. Consequently, the development of methods intended for the identification of mutation patterns present in cancer genomes and linking such patterns to specific mutagenic processes has become a research topic of broad interest as we discussed in our recent survey of the topic (ref 1). Importantly, approaches for de novo discovery of mutational signatures typically assume that the impacts of mutagenic processes are independent on each other and aim to define a set of signatures so that mutation counts observed in all samples can be optimally expressed as the sum of contributions of individual signatures weighted by the patient-specific signature activity (i.e. exposure). This assumption is an oversimplification. For example, deficient DNA mismatch repair processes contribute to cancer mutation catalogue by modifying the outcome of other mutagens. We developed a new method, called RepairSig based an advanced computational approach designed to close this gap. The method distinguishes two types of mutagenic processes: primary processes that affect the genomes independently from each other in an additive way, and non-additive secondary processes that act upon the mutations produced by primary processes. The former category includes UV light and smoking while the latter includes the deficient DNA mismatch repair mechanism (MMRd). Indeed, MMRd impacts the way in which errors introduced by the primary processes are repaired. Representing non-additive processes using additive models requires inferring a separate signature to explain each interaction and thus does not allow maintaining a clear correspondence between signatures and mutagenic processes. As a case in point, additive models identified several mutational signatures related to MMRd in breast cancer which, based on further analysis, correspond to such interactions. In contrast, RepairSig infers just one MMRd signature which corresponds to the experimentally validated signature. The paper describing these studies has been selected for presentation at RECOMB 2021 and accepted for publication in Cell Systems (ref 2). We also leverage the concept mutational to study the relationship of smoking and COV-19. To investigate the role of smoking in COVID-19 vulnerabilities, we utilized the large mutation and gene expression datasets of cancer and healthy tissues collected in TCGA. In particular, we focused on LUAD dataset (lung adenocarcinoma) as its etiology has many tangent points with COVID-19. Using a pathway level analysis integrating gene expression and mutational signatures, we examined the relationship of the exposure to smoking and other mutagens with biological processes in healthy tissues, aiming to understand how the exposure to these mutagens impact the infectivity of the virus and disease progression. Our preliminary results demonstrated that mutational signatures can be utilized to study the impact of mutagenic environmental factors on molecular pathways and cellular compositions of tissues by allowing a quantification of the strength of these mutagens. The analysis results with this approach are consistent with recent findings linking perturbations of these pathways to smoking but also provided additional novel insights. Overall, our study indicates that smoking changes expression of genes and pathways as well as the cell type composition in lung tissue in a way consistent with increasing the susceptibility to SARS-CoV-2 infection in lungs and contributing to progression to severe disease cases. A manuscript describing these studies is in preparation. Mutation heterogeneity underlines many aspects of phenotypic heterogeneity in cancer. Understanding such phenotype-genotype relationships is fundamental for the advance of personalized medicine. We developed a computational method, named NETPHIX (NETwork-to-PHenotype association with eXclusivity) to identify subnetworks of genes whose genetic alterations are associated with drug response or other continuous cancer phenotypes. Leveraging interaction information among genes and properties of cancer mutations such as mutual exclusivity, we formulate the problem as an integer linear program and solve it optimally to obtain a subnetwork of associated genes. Applied to a large-scale drug screening dataset, NETPHIX uncovered gene modules significantly associated with drug responses. Utilizing interaction information, NETPHIX modules are functionally coherent, and can thus provide important insights into drug action. In addition, we show that modules identified by NETPHIX together with their association patterns can be leveraged to suggest promissing drug combinations (ref 3).