Classification and annotations of cancer variants using structure-based methods

Cancer is characterized by heterogeneous alterations resulting in uncontrolled cell growth. Decoding the alterations, including mutations, and their intricacies has been significantly advanced by next-generation sequencing technologies. These technologies have accelerated the ability to identify an expansive array of genetic variants associated with cancer. However, the pace of variant discovery surpasses the capacity of clinical assessment, resulting in a discrepancy between the speed of discovery and clinical annotation. Indeed, genomic variation detection has resulted in numerous variants classified as Variants of Uncertain Significance (VUS) or associated with conflicting evidence. This disjunction is complicated as single genetic variants may elicit diverse clinical assessments. The challenge extends to understanding the pathogenicity of these variants. Despite the progress in bioinformatics and computational approaches, as reviewed in Manuscript I, accurately pinpointing the pathogenic nature of a genetic variant remains an open question in many instances. To navigate this complexity, this thesis aims to elucidate the intricate reasoning that a particular variant may be benign or pathogenic. The aim is to study how alterations affect the three-dimensional structure of proteins and subsequently provide valuable insights into their functional implications for the protein. This is investigated by delving into mutations within coding regions and assessing the structural consequences, explored in Manuscript II, where we present a protocol to analyze variants computationally and systematically, A Modular Structure-Based Framework for Genomic Variant (MAVISp). Where Manuscripts III, IV and V were part of the groundwork of this framework. The framework requires solved or predicted protein structures to model the alterations, presented in Manuscript III, an application note presenting PDBminer, a tool to ease the tedious task of choosing a structural model. Decoding the structural consequences requires multi-tool and -method interplay to understand why a variant may be associated with cancer and establish the specific mode of alteration leading to a pathway breakdown. One of the applied predictive modules is the stability alteration upon mutation, as introduced in Manuscripts IV and V, where we introduce high-throughput versions of FoldX, MutateX, and the Rosetta suite, RosettaDDGprediction. When the investigations rely on computational predictions, the tools should continually be validated against experimental findings, as shown in Manuscript VI, where we benchmark the stability predictors. To illustrate the utility of all this method and protocol development, we study p53 alterations from a structural perspective in Manuscript VII. In addition, we continue to use p53 as a model system to investigate the applicability of structural bioinformatics to broaden our understanding of mechanistic determinants in Manuscript VIII where we do a case study of compensatory mutations. Collectively, this thesis suggests that it may be possible to predict the culpable alternation as a mutational consequence by establishing how the protein, and its connected biological pathways, are changed, resulting in protein loss of function. This thesis suggests a unified framework to comprehensively understand the mutational landscape of cancer. Even though the statements in this thesis need further validation with more data, and the framework will benefit from continued implementation of methodologies, to continually account for more biological details, the Manuscripts of this thesis add novel insight into the field of structural cancer biology.