Efficient epistasis inference via higher-order covariance matrix factorization DOI Creative Commons
Kai Shimagaki, John P. Barton

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 14, 2024

Epistasis can profoundly influence evolutionary dynamics. Temporal genetic data, consisting of sequences sampled repeatedly from a population over time, provides unique resource to understand how epistasis shapes evolution. However, detecting epistatic interactions sequence data is technically challenging. Existing methods for identifying are computationally demanding, limiting their applicability real-world data. Here, we present novel computational method inferring that significantly reduces costs without sacrificing accuracy. We validated our approach in simulations and applied it study HIV-1 evolution multiple years set 16 individuals. There observed strong excess negative between beneficial mutations, especially mutations involved immune escape. Our general could be used characterize other large sets.

Language: Английский

Symmetry, gauge freedoms, and the interpretability of sequence-function relationships DOI Creative Commons

Anna Posfai,

David M. McCandlish, Justin B. Kinney

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: May 13, 2024

Quantitative models that describe how biological sequences encode functional activities are ubiquitous in modern biology. One important aspect of these is they commonly exhibit gauge freedoms, i.e., directions parameter space do not affect model predictions. In physics, freedoms arise when physical theories formulated ways respect fundamental symmetries. However, the connections sequence-function relationships have to symmetries sequence yet be systematically studied. Here we study a specific symmetry space: group position-specific character permutations. We find parameters transform under redundant irreducible matrix representations this group. Based on finding, an "embedding distillation" procedure enables analytic calculation number independent as well efficient computation sparse basis for freedoms. also transformation behavior affects interpretability. many (and possibly all) nontrivial models, ability interpret individual quantifying intrinsic allelic effects requires present. This finding establishes incompatibility between two distinct notions Our work thus advances understanding symmetries, and interpretability relationships.

Language: Английский

Citations

3
Symmetry, gauge freedoms, and the interpretability of sequence-function relationships DOI Creative Commons

Anna Posfai,

David M. McCandlish, Justin B. Kinney

et al.

Physical Review Research, Journal Year: 2025, Volume and Issue: 7(2)

Published: April 2, 2025

Quantitative models that describe how biological sequences encode functional activities are ubiquitous in modern biology. One important aspect of these is they commonly exhibit gauge freedoms, i.e., directions parameter space do not affect model predictions. In physics, freedoms arise when physical theories formulated ways respect fundamental symmetries. However, the connections sequence-function relationships have to symmetries sequence yet be systematically studied. this work we study a specific symmetry space: group position-specific character permutations. We find parameters transform under redundant irreducible matrix representations group. Based on finding, an “embedding distillation” procedure enables both analytic calculation number independent and efficient computation sparse basis for freedoms. also transformation behavior affects interpretability. many (and possibly all) nontrivial models, ability interpret individual as quantifying intrinsic allelic effects requires present. This finding establishes incompatibility between two distinct notions Our thus advances understanding symmetries, interpretability relationships. Published by American Physical Society 2025

Language: Английский

Citations

0
Efficient epistasis inference via higher-order covariance matrix factorization DOI Creative Commons
Kai Shimagaki, John P. Barton

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 14, 2024

Epistasis can profoundly influence evolutionary dynamics. Temporal genetic data, consisting of sequences sampled repeatedly from a population over time, provides unique resource to understand how epistasis shapes evolution. However, detecting epistatic interactions sequence data is technically challenging. Existing methods for identifying are computationally demanding, limiting their applicability real-world data. Here, we present novel computational method inferring that significantly reduces costs without sacrificing accuracy. We validated our approach in simulations and applied it study HIV-1 evolution multiple years set 16 individuals. There observed strong excess negative between beneficial mutations, especially mutations involved immune escape. Our general could be used characterize other large sets.

Language: Английский

Citations

2