Trevor Christensen

Research

Contents

I’m a PhD student in Peter Koo’s lab at Cold Spring Harbor Laboratory. I build machine-learning models that predict function directly from nucleic-acid sequence, and I focus on the data-centric side of that problem: which data to generate, and how to train on it efficiently.

What I work on

  • Data scaling laws for regulatory genomics. My thesis studies which sequence-generation strategies buy the most model improvement per experiment and per dollar, so that large training datasets can be designed deliberately rather than accumulated by default.
  • Efficient, reliable models. I’ve worked on knowledge distillation (producing models with ~10× faster inference at comparable accuracy) and uncertainty quantification, to make sequence-to-function models cheaper to deploy and easier to trust.
  • From sequence to therapeutics. I’m interested in applying these methods to RNA- and DNA-based therapeutics and vaccines, where experimental data is expensive and good design-of-experiments matters.

Background

Before CSHL I was a postbaccalaureate fellow at the National Cancer Institute, where I characterized pancreatic-cancer GWAS variants using RNA-seq and gene-regulatory-network inference. As an undergraduate at Claremont McKenna College, I engineered DNA polymerases to synthesize modified (2′-fluoro) nucleic acids relevant to aptamers and nucleic-acid therapeutics.

Publications

  • Uncertainty-aware genomic deep learning with knowledge distillation. Zhou, J., Rizzo, K., Christensen, T., Tang, Z., & Koo, P. K. (2026). npj Artificial Intelligence, 2, 3. doi.org/10.1038/s44387-025-00053-3

  • Characterization of a pancreatic cancer GWAS signal suggests PDX1 buffers stress in the exocrine pancreas. Hoskins, J. W., Christensen, T. A., et al. (2026). medRxiv (preprint). doi.org/10.64898/2026.04.13.26350790

  • Master regulator activity QTL protocol to implicate regulatory pathways potentially mediating GWAS signals using eQTL data. Hoskins, J. W., Christensen, T. A., & Amundadottir, L. T. (2023). STAR Protocols, 4(3), 102362. doi.org/10.1016/j.xpro.2023.102362

  • Mutant polymerases capable of 2′-fluoro-modified nucleic acid synthesis and amplification with improved accuracy. Christensen, T. A., Lee, K. Y., Gottlieb, S. Z. P., Carrier, M. B., & Leconte, A. M. (2022). RSC Chemical Biology, 3(8), 1044–1051. doi.org/10.1039/d2cb00064d

Elsewhere

Backlinks