Novel Uses for Machine Learning and Other Computational Methods for the Design and Interpretation of Genetic Microarrays

Abstract

It is clear that high-throughput techniques, such as rapid DNA sequencing and gene chips are changing the science of genetics. Hypothesis-driven science is now strongly complemented by these newer data-driven approaches. Over the course of the past decade, DNA microarrays, also known as gene chips, have come into prominence for genetic-level analysis throughout the life sciences. Using these microarrays, a scientist is able to perform hundreds of thousands of experiments on the surface of a single one-inch-by-one-inch wafer in the space of a single afternoon, generating more data than an army of researchers could have a generation ago. This potential flood of data brings many informatic challenges in both analysis and design. It is well understood that computer science will play a crucial role in their development and application. This thesis presents novel applications of machine learning and other computational methods to central tasks in highthroughput biology. These tasks include gene-chip design, detection of genomic variation, and the interpretation of gene-expression patterns.