Experimental and computational advances for studying the human genome with optical mapping

Many human geneticists were surprised by the discovery in 2004 that normal human genomes differ in structure as well as in sequence. Comprising up to 5% of the genome, these submicroscopic insertions, deletions, inversions and rearrangements represent a substantial source of genetic polymorphism and have been implicated in human evolution and disease. However, due to their large size and frequent association with repeated sequence, they remain poorly characterized by current methods.;This thesis addresses the discovery and characterization of genome structure variation in normal human genomes using Optical Mapping. Optical Mapping is a unique platform for analyzing genomes: it uses measurements of single molecules of DNA to infer a high-resolution genome-wide restriction map, whose representation of genome structure complements genome sequence to yield biological insight. These restriction maps are useful in a variety of genome analyses, including aiding in sequence assembly, probing cancer genomes for new therapeutic targets, and understanding normal human genetic variation.;The thesis begins with a careful optimization of many Optical Mapping protocols, with an eye towards improving throughput, consistency and data quality. Then, it describes the creation of genome-wide restriction maps of four normal human genomes, allowing us to analyze the structure of these genomes in unprecedented breadth and detail. The approach is validated by showing strong concordance with existing methods, while describing thousands of new variants with sizes ranging from kilobases to megabases of affected sequence.;The thesis concludes with the development of new analyses for Optical Mapping data sets based on a hidden Markov model (HMM). HMMs have found use in a variety of bioinformatics endeavors including gene finding, copy number analysis, secondary structure prediction and multiple sequence alignment. The best-studied problems on hidden Markov models (evaluation, decoding, and learning) translate directly to common tasks in analyzing Optical Mapping data and provide a jumping-off point for addressing more interesting problems including restriction map refinement and haplotype discernment.