Scalable saturation mutagenesis reveals gene regulatory architecture and rare variant effects
- 1Calico Life Sciences LLC
Abstract
Long-context sequence-to-function models enable nucleotide-resolution prediction of regulatory variant effects, motivating comprehensive mutational interrogation across broad genomic contexts. Yet conventional in silico saturation mutagenesis (ISM) scores mutations one at a time, requiring millions of model evaluations for a single gene and billions to trillions at genome scales. Here, we introduce Multi-ISM, a scalable framework that reformulates ISM as a sparse recovery problem. Multi-ISM generates mutational maps using 45-fold fewer model evaluations than exhaustive single-variant ISM and matches or exceeds its accuracy on variant-effect benchmarks. Multi-ISM is architecture-agnostic and transfers across large-scale sequence-to-function models. We applied Multi-ISM to 5,000 protein-coding genes, including 3,317 OMIM disease genes, generating base-pair-resolution, tissue-resolved attribution maps across 500-kb windows. These maps supported enhancer--gene prioritization and identification of cell-type-specific regulatory elements. Gene-level summaries of the maps captured regulatory complexity and showed that more constrained genes had smaller predicted mutational effects. Aggregating Multi-ISM predictions into gene-level rare-variant burdens improved personalized expression prediction over a common-variant elastic net, with the largest gains at expression outliers. Multi-ISM makes nucleotide-resolution interpretation of long-context sequence models a routine computation rather than a dedicated effort, so that new architectures, functional readouts, and cellular contexts can be mapped as they appear.
Related
- Cardiomyocyte vulnerability to lamin polymer disruption revealed by saturation mutagenesis
- Hairpin Architecture and Extended Sequence Context Shape APOBEC3 Mutagenesis in Monkeypox Virus at Different Genomic Scales
- Deconvolution of HIV-1 Mutational Signatures Reveals Dominant and Donor-Specific APOBEC3-Associated Mutagenesis Across Anatomical Compartments
- Temporal control of mitochondrial mutagenesis reveals the fate of mtDNA mutations with age
- Functional decoding reveals a hidden regulatory layer of the Salmonella transcriptome during infection
BibTeX
@misc{yuan2026scalable,
title = {{Scalable saturation mutagenesis reveals gene regulatory architecture and rare variant effects}},
author = {H. Yuan and X. Huang and B. Auerbach and J. Linder and D. Srivastava and D. R. Kelley},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.30.755794},
url = {https://www.biorxiv.org/content/10.64898/2026.09.30.755794v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Scalable saturation mutagenesis reveals gene regulatory architecture and rare variant effects
AU - H. Yuan
AU - X. Huang
AU - B. Auerbach
AU - J. Linder
AU - D. Srivastava
AU - D. R. Kelley
PY - 2026
DA - 2026/09/30
PB - bioRxiv
DO - 10.64898/2026.09.30.755794
UR - https://www.biorxiv.org/content/10.64898/2026.09.30.755794v1
AB - Long-context sequence-to-function models enable nucleotide-resolution prediction of regulatory variant effects, motivating comprehensive mutational interrogation across broad genomic contexts. Yet conventional in silico saturation mutagenesis (ISM) scores mutations one at a time, requiring millions of model evaluations for a single gene and billions to trillions at genome scales. Here, we introduce Multi-ISM, a scalable framework that reformulates ISM as a sparse recovery problem. Multi-ISM generates mutational maps using 45-fold fewer model evaluations than exhaustive single-variant ISM and matches or exceeds its accuracy on variant-effect benchmarks. Multi-ISM is architecture-agnostic and transfers across large-scale sequence-to-function models. We applied Multi-ISM to 5,000 protein-coding genes, including 3,317 OMIM disease genes, generating base-pair-resolution, tissue-resolved attribution maps across 500-kb windows. These maps supported enhancer--gene prioritization and identification of cell-type-specific regulatory elements. Gene-level summaries of the maps captured regulatory complexity and showed that more constrained genes had smaller predicted mutational effects. Aggregating Multi-ISM predictions into gene-level rare-variant burdens improved personalized expression prediction over a common-variant elastic net, with the largest gains at expression outliers. Multi-ISM makes nucleotide-resolution interpretation of long-context sequence models a routine computation rather than a dedicated effort, so that new architectures, functional readouts, and cellular contexts can be mapped as they appear.
ER -