Kramer, A. (BMEB) – Scalable phylo-pangenomics

The COVID-19 pandemic generated genomic data at unprecedented scale, with tens of millions of SARS-CoV-2 genomes deposited in public repositories and thousands of new sequences added each day. This dissertation develops methods for analyzing genomic datasets at this scale, unified by the idea that encoding genomes according to their evolutionary relationships can make otherwise intractable computations practical. First, I evaluate online phylogenetic inference, in which new genomes are continuously added to an existing tree, and compare parsimony-based methods with maximum-likelihood approaches under pandemic time constraints. For densely sampled SARS-CoV-2 genomes, online inference with UShER and matOptimize produces trees comparable to established maximum-likelihood methods while requiring orders of magnitude less time and memory. I then develop tools that make phylogenies containing millions of genomes useful for downstream analysis and visualization. ShUShER enables privacy-preserving phylogenetic placement within a web browser, allowing laboratories to analyze sensitive sequences without transmitting them to an external server. Treenome Browser co-visualizes the genomic variation of millions of samples alongside their phylogenetic relationships by operating directly on a compressed mutation-annotated tree. Finally, I describe Panmap, which uses Pangenome Mutation-Annotated Networks (PanMANs) to place, align, and genotype sequencing reads and to estimate haplotype abundances against reference collections containing up to millions of genomes. Panmap produces indexes hundreds of times smaller than graph-based alternatives, improves genome reconstruction over single-reference workflows at low coverage, and supports applications ranging from pathogen genome assembly to ancient environmental DNA analysis. Together, these results show that evolutionary history can serve not only as an object of inference but as a scalable computational infrastructure for phylogenomic and pangenomic analyses as genomic datasets continue to grow.

Event Host: Alexander Kramer, Ph.D. Candidate, Biomolecular Engineering & Bioinformatics

Advisor: Russell Corbett-Detig

Zoom: https://ucsc.zoom.us/j/91364025182?pwd=Tq95CuaBqrePatRjopy6uJ9bbjsrIH.1

Passcode: 318268