Topic

Genomics

Large-scale sequencing data and the statistical and machine-learning models that turn genomes into biological and clinical insight.

1 episode

About Genomics on Interpretable

Genomics turns sequences into information about biology, and this topic covers what the podcast has said about doing that at scale with models people can read. Its episode is Episode 002, which looks at a 2022 Science study of 6.4 million publicly available SARS-CoV-2 genomes by Jacob E. Lemieux and colleagues, including lead author Fritz Obermeyer at the Broad Institute of MIT and Harvard.

The study's model, PyR0, is a hierarchical Bayesian multinomial logistic regression, written in the Pyro probabilistic programming language. In the talk, Lemieux explains the model formulation and how it is written in Pyro, shows results on mutation-specific fitness, and asks which lineages are most volatile and what is driving fitness. The later chapters compare PyR0 with other approaches, including the Lee and Barton approach and Bayesian viral allele selection. The central idea is that a genome is read mutation by mutation: each substitution receives an estimate of its effect on growth, which a laboratory can then test. The episode and its paper, with open-access and preprint versions, are listed below.

Episode

References

  1. Fritz Obermeyer, Martin Jankowiak, Nikolaos Barkas, Stephen F. Schaffner, Jesse D. Pyle, Leonid Yurkovetskiy, Matteo Bosso, et al. (2022). Analysis of 6.4 million SARS-CoV-2 genomes identifies mutations associated with fitness(opens in a new tab). Science, 376 (6599): 1327–1332. https://doi.org/10.1126/science.abm1208. Discussed in Which Mutations Make a Virus Spread? Reading 6.4 Million Genomes.