Predicting genome-wide functional constraints with GPN-Star
8 affiliations
- 1Department of Statistics, University of California, Berkeley, CA, USA
- 2Computer Science Division, University of California, Berkeley, CA, USA
- 3The Jackson Laboratory, Bar Harbor, ME, USA
- 4Division of Computational Genomics and Systems Genetics, German Cancer Research Center (DKFZ), Heidelberg, Germany
- 5Center for Computational Biology, University of California, Berkeley, CA, USA. [email protected]
- 6Computer Science Division, University of California, Berkeley, CA, USA. [email protected]
- 7Department of Statistics, University of California, Berkeley, CA, USA. [email protected]
- 8Innovative Genomics Institute, University of California, Berkeley, CA, USA. [email protected]
Abstract
Genomic language models have emerged as a powerful approach for learning genome-wide functional constraints directly from DNA sequences 1 . However, standard genomic language models adapted from natural language processing often require large model sizes and computational resources, yet still fall short of classical evolutionary models in predictive tasks 2–4 . Here we introduce a genomic pretrained network with species tree and alignment representations (GPN-Star), which is a biologically grounded genomic language model featuring a phylogeny-aware architecture that leverages whole-genome alignments and species trees to model evolutionary relationships explicitly. Trained on alignments spanning vertebrate, mammal and primate evolutionary timescales, GPN-Star achieves state-of-the-art performance across a wide range of variant effect prediction tasks in both coding and non-coding regions of the human genome. Analyses across timescales show task-dependent advantages of modelling more recent versus deeper evolution. To demonstrate its potential to advance human genetics, we show that GPN-Star substantially outperforms previous methods in prioritizing pathogenic and fine-mapped genome-wide association study variants, yields strong enrichments of complex trait heritability and improves power in rare variant association testing 5 . Extending beyond humans, we train GPN-Star for five model organisms— Mus musculus , Gallus gallus , Drosophila melanogaster , Caenorhabditis elegans and Arabidopsis thaliana —demonstrating the robustness and generalizability of the framework. Taken together, these results position GPN-Star as a scalable, powerful and flexible tool for genome interpretation, well suited to leverage the growing abundance of comparative genomics data.
Related
- Genome-Wide Identification of the Two-Pore Potassium Channel Gene Family in Sugar Beet (Beta vulgaris L.) and Functional Analysis of BvTPK2 in Salt Tolerance
- Genome-Wide Identification and Functional Validation of the Cold-Tolerance WRKY Transcription Factor in Sweetpotato
- Genome-Wide Identification and Functional Characterization of the Carboxylesterase (CXE) Gene Family in Melon (Cucumis melo L.): CmCXE7 as a Cold-Responsive Regulator of Ester Metabolism
- Genome-Wide Characterization of the Wheat Pum Gene Family and Heterologous Functional Analysis of TaPum-III-5b in Arabidopsis Under Salt Stress
- Genome-wide identification of the BES1 gene family and functional study of VcBES1.3 under salt stress in blueberry (Vaccinium corymbosum)
BibTeX
@article{ye2026predicting,
title = {{Predicting genome-wide functional constraints with GPN-Star}},
author = {Chengzhong Ye and Gonzalo Benegas and Carlos Albors and Jianan Canal Li and Sebastian Prillo and Peter D. Fields and Brian Clarke and Yun S. Song},
journal = {Nature},
year = {2026},
month = sep,
doi = {10.1038/s41586-026-11005-5},
url = {https://doi.org/10.1038/s41586-026-11005-5},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Predicting genome-wide functional constraints with GPN-Star
AU - Chengzhong Ye
AU - Gonzalo Benegas
AU - Carlos Albors
AU - Jianan Canal Li
AU - Sebastian Prillo
AU - Peter D. Fields
AU - Brian Clarke
AU - Yun S. Song
PY - 2026
DA - 2026/09/09
JO - Nature
DO - 10.1038/s41586-026-11005-5
UR - https://doi.org/10.1038/s41586-026-11005-5
AB - Genomic language models have emerged as a powerful approach for learning genome-wide functional constraints directly from DNA sequences 1 . However, standard genomic language models adapted from natural language processing often require large model sizes and computational resources, yet still fall short of classical evolutionary models in predictive tasks 2–4 . Here we introduce a genomic pretrained network with species tree and alignment representations (GPN-Star), which is a biologically grounded genomic language model featuring a phylogeny-aware architecture that leverages whole-genome alignments and species trees to model evolutionary relationships explicitly. Trained on alignments spanning vertebrate, mammal and primate evolutionary timescales, GPN-Star achieves state-of-the-art performance across a wide range of variant effect prediction tasks in both coding and non-coding regions of the human genome. Analyses across timescales show task-dependent advantages of modelling more recent versus deeper evolution. To demonstrate its potential to advance human genetics, we show that GPN-Star substantially outperforms previous methods in prioritizing pathogenic and fine-mapped genome-wide association study variants, yields strong enrichments of complex trait heritability and improves power in rare variant association testing 5 . Extending beyond humans, we train GPN-Star for five model organisms— Mus musculus , Gallus gallus , Drosophila melanogaster , Caenorhabditis elegans and Arabidopsis thaliana —demonstrating the robustness and generalizability of the framework. Taken together, these results position GPN-Star as a scalable, powerful and flexible tool for genome interpretation, well suited to leverage the growing abundance of comparative genomics data.
SN - 0028-0836
SN - 1476-4687
ER -