Mastoras, M. (BMEB) – Polishing genome assemblies and leveraging their improved quality to study centromere variation

A complete and accurate genome reconstruction serves as the foundation for studying an organism’s biology and the mechanisms underlying disease. It is particularly critical for reference genomes, which provide a universal coordinate system for downstream genomic analysis. Errors or missing sequences in a reference genome create bias in all of the studies built on top of them. The Human Pangenome Reference Consortium (HPRC) seeks to address this bias by transitioning the field to a pangenome reference, a graph based collection of many genome assemblies, providing a better representation of variation in the human population. Removing errors in the HPRC assemblies is critical to ensure the pangenome serves as a robust standard for genomic variant discovery. In the first part of my thesis, I improve the base level accuracy of the HPRC release 2 assemblies (HPRC2) with a machine learning model for assembly polishing called DeepPolisher. Next, I make additional contributions to reference-based genomic analysis by polishing reference genomes of other model organisms, and helping to develop a new method for de-novo assembly and variant calling from a single-flow cell nanopore sequencing protocol. Finally, I take advantage of the highly accurate, near complete assemblies from HPRC2 that I improved with DeepPolisher to study a region only recently made accessible to genomics analysis: the human centromere. I apply the tool Centrolign, the first ever multiple-sequence-aligner for centromeres to the HPRC2 assemblies, establishing precise estimates of mutation rates and spatial variation patterns across centromeric arrays.

Event Host: Mira Mastoras, Ph.D. Candidate, Biomolecular Engineering & Bioinformatics

Advisor: Benedict Paten

Zoom: https://ucsc.zoom.us/j/98960011053?pwd=mFKCwtSIhvT5FvEURbE85lRbbMvzuo.1

Passcode: 118488