Search

High-throughput sequencing methods now provide extensive data on disease-related human genetic variants. New methods are required to maximally utilize these data for enhanced understanding and treatment of human diseases. This dissertation describes my work in addressing three aspects of this challenge: Determining disease-causative variants; representing mechanisms by which genetic variant(s) cause disease phenotypes; and quantitatively analyzing genetic disease mechanisms. First, I developed a variant prioritization algorithm, VarP, and objectively tested it in CAGI (Critical Assessment of Genome Interpretation). It was ranked best in the CAGI challenge on interpreting panel sequencing data for 106 patients, determining which disease class each patient has and the corresponding causative variant(s). VarP correctly identified the disease class for 36 cases, including 10 where the original clinical pipeline failed, and found seven cases with strong evidence of an alternative disease to that tested. Over-reliance on pathogenicity annotations in the HGMD mutation database led to several incorrect cases. Post analysis showed that protein structure data could have helped to interpret the impact of many prioritized missense variants. Next, I co-developed and implemented MecCog, a web-based graphical framework to represent mechanisms by which genetic variants cause disease phenotypes. A MecCog mechanism schema displays the propagation of system perturbations across stages of biological organization, using graphical notations to symbolize perturbed entities and activities, knowledge gaps, ambiguities and uncertainties, and hyperlinked evidence. The web platform enables a user to construct, store, publish, browse, query, and comment on schemas. MecCog facilitates better comprehension of disease mechanisms, identification of critical unanswered questions on causal relationships, and possible new sites of therapeutic intervention. Finally, I developed a framework to quantitatively represent and analyze mechanisms relating genetic variants to complex trait disease. It involves generating a computable circuit from MecCog schemas by assigning node functions and parameters to represent the behavior of the schema components. I demonstrate that such a circuit can be used to analyze the effect size of a variant contributing to disease risk as a function of the genetic background in an individual and the extent to which epistatic effects may be masked in population averages. I also show that the circuit functions and parameters can be learned in a data-driven manner using a hybrid neural network approach.

Mechanism is a widely used concept in biology: in 2017 more than 10% of PubMed abstracts used the term. Thus, searching for and reasoning about mechanisms is fundamental to much of biomedical research, but until now there has been almost no computational infrastructure for this purpose. Recent work in the philosophy of science has explored the central role that the search for mechanistic accounts of biological phenomena plays in biomedical research, providing a conceptual basis for representing and analyzing biological mechanism. The foundational categories for components of mechanisms - entities and activities - guide the development of general, abstract types of biological mechanism parts. Building on that analysis, we have developed a formal framework for describing and representing biological mechanism, MecCog, and applied it to describing mechanisms underlying human genetic disease. Mechanisms are depicted using a graphical notation. Key features are assignment of mechanism components to stages of biological organization and classes; visual representation of uncertainty, ignorance, and ambiguity; and tight integration with literature sources. The MecCog framework facilitates analysis of many aspects of disease mechanism, including the prioritization of future experiments, probing of gene-drug and gene-environment interactions, identification of possible new drug targets, personalized drug choice, analysis of non-linear interactions between relevant genetic loci, and classification of disease based on mechanism.

Precision medicine aims to predict a patient's disease risk and best therapeutic options by using that individual's genetic sequencing data. The Critical Assessment of Genome Interpretation (CAGI) is a community experiment consisting of genotype–phenotype prediction challenges; participants build models, undergo assessment, and share key findings. For CAGI 4, three challenges involved using exome-sequencing data: Crohn's disease, bipolar disorder, and warfarin dosing. Previous CAGI challenges included prior versions of the Crohn's disease challenge. Here, we discuss the range of techniques used for phenotype prediction as well as the methods used for assessing predictive models. Additionally, we outline some of the difficulties associated with making predictions and evaluating them. The lessons learned from the exome challenges can be applied to both research and clinical efforts to improve phenotype prediction from genotype. In addition, these challenges serve as a vehicle for sharing clinical and research exome data in a secure manner with scientists who have a broad range of expertise, contributing to a collaborative effort to advance our understanding of genotype–phenotype relationships.