Cole Williams

About

Postdoctoral Researcher · UCSF

I am a computational geneticist and postdoctoral researcher at UCSF in the Ziv Lab. My PhD at Brown University with Sohini Ramachandran focused on population genetics methods, including algorithms for pedigree inference and IBD-based genome phasing.

At UCSF, my work centers on cancer genomics and immunogenomics. I study how germline genetic variation shapes the tumor immune microenvironment — including pan-cancer GWAS of transposable element expression in tumors and its relationship to immune activation. I also work on germline predictors of immune-related adverse events in patients receiving cancer immunotherapy, and on polygenic risk score methodology for breast cancer and immunotherapy-related toxicities.

Google Scholar GitHub CV

Software

PONDEROSA

Genealogical relationship inference through identity-by-descent segments. Designed for large-scale datasets and endogamous populations.

IBD · Relatedness · Pedigrees

GitHub

HAPTIC

A tool for inter-chromosomal haplotype phasing.

Phasing · Haplotypes

GitHub

IBD Simulator

Simulation pipeline for simulating realistic identity-by-descent segments in Wright-Fisher populations. Uses msprime as the core simulation engine.

IBD · Simulation

GitHub

Tools

C4 Toolkit

A Python library for genetics research, originally developed for UCSF's C4 cluster but adaptable to any workspace on any HPC or computing environment. A single catalog file is used to find, load, and merge lab datasets, linking clinical, exome, and array IDs to one patient ID. HLA utilities parse and harmonize allele calls across formats and resolutions, so there's no more ad hoc string wrangling. Also includes support for association testing, survival analyses, and fine-mapping.

Data Management · HLA · Association Testing

Documentation

TE Co-expression Explorer

An interactive map of transposable element (TE) subfamily co-expression across TCGA tumors. Each point is a TE subfamily, placed by principal component analysis so that co-expressed TEs sit close together. Choose a cancer type and components to plot, color by TE class or family, and search for TEs by name.

Transposable Elements · TCGA · PCA

Open Explorer

Publications

Phasing millions of samples achieves near perfect accuracy, enabling parent-of-origin analyses.
CM Williams, J O'Connell, E Jewett, WA Freyman, 23andMe Research Team, CR Gignoux, S Ramachandran, AL Williams. HGG Advances, 6(4):100479. (2025)

A rapid accurate approach to inferring pedigrees in endogamous populations.
CM Williams, BA Scelza, SD Slack, N Font-Porterias, DR Al-Hindi, RA Mathias, H Watson, KC Barnes, E Lange, RK Johnson, CR Gignoux, S Ramachandran, BM Henn. Genetics, 230(4):iyaf094. (2025)

Principles and Policy Recommendations for Comprehensive Genetic Data Governance.
V Ramanan*, R Vinod*, C Williams*, S Ramachandran, S Venkatasubramanian. *Equal contribution. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (AIES-25), Vol. 8(3). (2025)

Social mates dynamically coordinate aggressive behavior to produce strategic territorial defense.
N Moody, CM Williams, S Ramachandran, M Fuxjager. PLOS Computational Biology. (2025)

Increased homozygosity due to endogamy results in fitness consequences in a human population.
NA Swinford, SP Prall, S Gopalan, CM Williams, J Sheehama, BA Scelza. PNAS, 120(43), e2309552120. (2023)

Building a vertically integrated genomic learning health system: The biobank at the Colorado Center for Personalized Medicine.
LK Wiley, JA Shortt, ER Roberts....CM Williams et al. The American Journal of Human Genetics, 111(1), 11–23. (2022)

Hunter-gatherer genomes reveal diverse demographic trajectories during the rise of farming in Eastern Africa.
S Gopalan, REW Berl, JW Myrick...CM Williams et al. Current Biology, 32(8), 1852–1860. (2022)

Unpublished Work

Genetic variation visualization with PCA

An interactive tutorial exploring PCA to visualize genetic variation across the 1000 Genomes Project. Data available on request.

Separation of 1000 Genomes Project samples on PC1 as the number of markers increases.

Open in Google Colab

PoNNderosa coming soon

A new version of PONDEROSA for unphased data using convolutional neural networks (CNN) to learn IBD sharing patterns.

Contact

cole.williams@ucsf.edu