Mendel, a foundation model of human genetic variation, prioritizes regulatory variants and improves gene expression prediction
Abstract
Human genome-wide association studies have identified hundreds of thousands of variant trait associations, but interpreting their mechanistic consequences at allele resolution remains a central bottleneck. DNA foundation models pretrained on genomic sequence corpora learn features predictive of regulatory activity, yet their training objectives are dominated by the conserved background common to all individuals, rather than the sparse variant sites at which human genetic variation is concentrated. Here we introduce Mendel, a DNA foundation model trained from scratch on human whole genome sequences using a variant-centric objective that concentrates supervision on allelic events within reference aligned sequence context. On held-out individuals, Mendel attains 0.984 micro-averaged SNV allele-prediction accuracy, compared with 0.311 for Evo 2. Mendel residual difficulty is selectively enriched at cis-eQTL variants, prioritizes fine-mapped causal variants (AUROC 0.731), where Evo 2 residuals performed near chance, and when converted into a cis-expression prior for modelling, improves prediction relative to an unweighted baseline in all 50 GTEx tissues.
Related
- VOICE: A Vision-Omics Foundation Model Integrating Direct and Retrieval-Based Prediction of In-situ Single-Cell Gene Expression
- REN-former prioritizes candidate regulators of kidney disease-state transitions through single-cell foundation modeling and human genetics
- Limitations of Genomic Foundation Models for Decoding Regulatory Mechanisms in ALS
- Where single-cell foundation models help: a benchmark of annotation, perturbation and regulatory inference against simple baselines
- DAPE-PD: Diffusion adaptation of genomic foundation model enables tailored-generation of cis-regulatory DNA
BibTeX
@misc{wu2026mendel,
title = {{Mendel, a foundation model of human genetic variation, prioritizes regulatory variants and improves gene expression prediction}},
author = {Chong Wu and Arham Salman and Haonan Feng and Peng Wei and Ryan Sun and Lang Wu and Wei Pan},
howpublished = {Research Square},
year = {2026},
month = oct,
doi = {10.21203/rs.3.rs-11197302/v1},
url = {https://doi.org/10.21203/rs.3.rs-11197302/v1},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Mendel, a foundation model of human genetic variation, prioritizes regulatory variants and improves gene expression prediction
AU - Chong Wu
AU - Arham Salman
AU - Haonan Feng
AU - Peng Wei
AU - Ryan Sun
AU - Lang Wu
AU - Wei Pan
PY - 2026
DA - 2026/10/26
PB - Research Square
DO - 10.21203/rs.3.rs-11197302/v1
UR - https://doi.org/10.21203/rs.3.rs-11197302/v1
AB - Human genome-wide association studies have identified hundreds of thousands of variant trait associations, but interpreting their mechanistic consequences at allele resolution remains a central bottleneck. DNA foundation models pretrained on genomic sequence corpora learn features predictive of regulatory activity, yet their training objectives are dominated by the conserved background common to all individuals, rather than the sparse variant sites at which human genetic variation is concentrated. Here we introduce Mendel, a DNA foundation model trained from scratch on human whole genome sequences using a variant-centric objective that concentrates supervision on allelic events within reference aligned sequence context. On held-out individuals, Mendel attains 0.984 micro-averaged SNV allele-prediction accuracy, compared with 0.311 for Evo 2. Mendel residual difficulty is selectively enriched at cis-eQTL variants, prioritizes fine-mapped causal variants (AUROC 0.731), where Evo 2 residuals performed near chance, and when converted into a cis-expression prior for modelling, improves prediction relative to an unweighted baseline in all 50 GTEx tissues.
SN - 2693-5015
ER -