Sequence-to-function models are fine-tuned and extended to score noncoding variants at…

Published papers

• A framework for genome-wide characterization of DNA methylation and variant effect prediction.

  Nature genetics2026-11-03Weiguang Mao et al.

  Presents a framework for genome-wide DNA methylation characterization, with a harmonized compendium of over 90 tissues and cell types and Wreath, a deep…

• XGBoost-SHAP analysis prioritizes lower-frequency variants predicted to affect HNF4A binding motifs in pancreatic progenitors.

  PLoS genetics2026-11-04Kelly A Frazer et al.

  XGBoost-SHAP framework on snATAC-seq from iPSC-derived pancreatic progenitors prioritizes noncoding SNPs affecting HNF4A binding, outperforming linear models.

Preprints

• Localizing predictive concepts encoded in neural representations via hierarchical testing

  arXiv2026-10-29Viktoriia Huryn et al.

  CUTTIE clusters neural network units into a tree and tests clusters top-down with conditional independence tests, controlling family-wise error to localize…

• Scooby-2: fine-tuning AlphaGenome for single-cell genomic track prediction

  bioRxiv2026-10-29Hingerl, J. C. et al.

  Scooby-2 fine-tunes AlphaGenome with a cell-specific decoder to predict single-cell RNA-seq and chromatin accessibility tracks from DNA sequence, outperforming…

• Unsupervised prioritization of gene-regulatory variants with isolation forests

  bioRxiv2026-11-03Schmidt, A. W. et al.

  eQIF applies unsupervised isolation forests to integrate Enformer predictions into one gene-regulatory score per common variant, prioritizing eQTL and GWAS…

• FiberBrowser: a unified framework for single-molecule genomics

  bioRxiv2026-10-31Tullius, T. W. et al.

  FiberBrowser is a unified platform and HMM inference engine (FiberHMM) that converts diverse single-molecule footprinting datasets into per-molecule chromatin…

• ChAI-Net learns a regulatory code for chromatin co-accessibility

  bioRxiv2026-10-31Lee, J. et al.

  ChAI-Net, a neural network predicting locus-specific chromatin co-accessibility from DNA sequence using single-molecule footprinting data in mouse embryonic…

• Predicting single-molecule chromatin distribution from DNA sequence

  bioRxiv2026-11-04Wang, R. et al.

  Introduces S2d, a sequence-to-distribution framework, and Mola, a model predicting single-molecule chromatin accessibility from DNA sequence, revealing…

• Decoding Cell-type Specific Gene Regulation with CellReg3D by Integrating DNA Large Language Models and 3D Genome Structure

  bioRxiv2026-11-04Wang, X. et al.

  CellReg3D is a multimodal deep learning framework combining DNA language model sequence embeddings with Hi-C 3D chromatin interactions to predict…

• Synthetic MPRA perturbation supervision improves variant-effect prediction in sequence-to-function models

  bioRxiv2026-11-03Wang, L. et al.

  Fine-tunes Borzoi sequence-to-function model on synthetic MPRA variant-effect labels, raising CAGI5 and eQTL Pearson correlations while preserving pre-trained…

Related topics

• Deep Learning Regulatory Prediction
• Cell Free Dna Methylation
• Sequence To Function Prediction Models

Topic history