Computational contributions towards scalable and efficient genome-wide association methodology

摘要: Genome-wide association studies are experiments designed to find the genetic bases of physical traits: for example, markers correlated with disease status by comparing DNA healthy individuals affecteds. Over past two decades, an exponential increase in resolution DNA-testing technology coupled a substantial drop their cost have allowed us amass huge and potentially invaluable datasets conduct such comparative studies. For many common diseases, as large hundred thousand exist, each tested at million(s) (called SNPs) across genome. Despite this treasure trove, so far only small fraction underlying most diseases been identified. Simply stated—our ability predict phenotype (disease status) from person's constitution is still very limited today, even traits that we know be heritable one's parents (e.g. height, diabetes, cardiac health). As result, genetics today often lags behind conventional indicators like family history terms its predictive power. To borrow popular metaphor astronomy, veritable "dark matter" perceivable but un-locatable signal has come known missing heritability. This thesis will present my research contributions hotly pursued scientific hypotheses aim close gap: (1) gene-gene interactions, (2) ultra-rare variants—both which not yet widely tested. First, I discuss challenges made interaction testing difficult, novel approximate statistic measure interaction. This can exploited Monte-Carlo randomization scheme, making exhaustive search through trillions potential interactions tractable using ordinary desktop computers. A software implementation our algorithm found reproducible between SNPs calcium channel genes Bipolar Disorder. Next, functional enrichment pipeline subsequently developed identify sets interacting disease. Lastly, talk about application coding theory cost-efficient measurement variation (sometimes, rare just one individual carrying mutation entire population).