gRely: Relyability for genome trained sequence-to-expression models

Abstract

Sequence-to-function (S2F) models predict molecular phenotypes from DNA sequence and are increasingly applied to variant effect prediction (VEP), where the goal is to quantify how genetic variants alter gene expression. However, S2F model predictions are not uniformly reliable: accuracy varies substantially across variants, genes, and tissues, and current practice relies on crude magnitude thresholding to enrich for trustworthy predictions, which discards the majority of variants where S2F models could still provide signal. We developed gRely, a meta-modeling framework that estimates the probability that a given Borzoi VEP correctly predicts eQTL direction, using 1,121 features derived from the target variant, gene, and model outputs. On held-out tissues, gRely achieves a mean average precision of 0.885 (random baseline 0.744). Critically, within the low-magnitude regime where thresholding fails entirely, gRely identifies a high-confidence subset with 76% accuracy compared to a 58% baseline, recovering reliable predictions that magnitude filtering would discard. Interpretation via SHAP reveals that in this low-magnitude regime, gene expression level and cross-replicate signal concentration replace VEP magnitude as the primary discriminators of reliability. gRely is the first framework to provide per-prediction confidence scores for S2F model VEPs, and generalizes across architectures, producing consistent improvements on AlphaGenome predictions. By making reliability quantifiable, gRely enables principled filtering rather than blanket thresholding, and marks a step toward trustworthy deployment of S2F models in genomic research and clinical applications.

Fine-tuning sequence-to-expression models on personal genome and transcriptome data
Ruchir Rastogi, Aniketh Janardhan Reddy, Ryan Chung +1
Genome biology25 May 2026Peer-reviewedCC BY21 citedGenetics & genomicsOpenAlex
S2F-Agent: Harnessing sequence-to-function models for verifiable genome interpretation
J. Li, T. Qin, J. G. Li +1
bioRxiv13 May 2026PreprintAll rights reservedFull textBioinformatics & computational biologybioRxiv
Complete genome sequence of Comamonas testosteroni TA441 and comparative analysis of steroid degradation gene clusters across Proteobacteria and Actinomycetota
M. Horinouchi, N. Kobayashi-Yoshida, H. Nojiri +1
bioRxiv18 Sep 2026PreprintAll rights reservedFull textMicrobiologybioRxiv
The complete genome sequence of a poplar, Populus davidiana.
Song Chen, Xinyu Wang, Yunxia Wei +13
Horticulture research1 Jun 2026Peer-reviewedCC BYFull text4 citedPlant biologyEurope PMCPMC
Tracing genome size dynamics in sharks and rays with inclusive sequence analysis by the Squalomix Consortium
Shigehiro Kuraku, Yawako W. Kawaguchi, Taiki Niwa +10
Genome Research16 Jun 2026PreprintLicence not statedFull text5 citedPlant biologyOpenAlex

Cite

BibTeX RIS

BibTeX

@misc{rafi2026grely,
  title = {{gRely: Relyability for genome trained sequence-to-expression models}},
  author = {A. M. Rafi and G. Eraslan and K. Fletez-Brant},
  howpublished = {bioRxiv},
  year = {2026},
  month = may,
  doi = {10.64898/2026.05.23.727431},
  url = {https://www.biorxiv.org/content/10.64898/2026.05.23.727431v2},
  note = {CC BY},
}

RIS

TY  - UNPB
TI  - gRely: Relyability for genome trained sequence-to-expression models
AU  - A. M. Rafi
AU  - G. Eraslan
AU  - K. Fletez-Brant
PY  - 2026
DA  - 2026/05/23
PB  - bioRxiv
DO  - 10.64898/2026.05.23.727431
UR  - https://www.biorxiv.org/content/10.64898/2026.05.23.727431v2
AB  - Sequence-to-function (S2F) models predict molecular phenotypes from DNA sequence and are increasingly applied to variant effect prediction (VEP), where the goal is to quantify how genetic variants alter gene expression. However, S2F model predictions are not uniformly reliable: accuracy varies substantially across variants, genes, and tissues, and current practice relies on crude magnitude thresholding to enrich for trustworthy predictions, which discards the majority of variants where S2F models could still provide signal. We developed gRely, a meta-modeling framework that estimates the probability that a given Borzoi VEP correctly predicts eQTL direction, using 1,121 features derived from the target variant, gene, and model outputs. On held-out tissues, gRely achieves a mean average precision of 0.885 (random baseline 0.744). Critically, within the low-magnitude regime where thresholding fails entirely, gRely identifies a high-confidence subset with 76% accuracy compared to a 58% baseline, recovering reliable predictions that magnitude filtering would discard. Interpretation via SHAP reveals that in this low-magnitude regime, gene expression level and cross-replicate signal concentration replace VEP magnitude as the primary discriminators of reliability. gRely is the first framework to provide per-prediction confidence scores for S2F model VEPs, and generalizes across architectures, producing consistent improvements on AlphaGenome predictions. By making reliability quantifiable, gRely enables principled filtering rather than blanket thresholding, and marks a step toward trustworthy deployment of S2F models in genomic research and clinical applications.
ER  -

Abstract

Related

Cite