Resources

Models, data and code

Everything we release is open. Use it, build on it, and tell us what you find.

DeepVRegulome models

464 fine-tuned DNABERT models for transcription factor binding, histone marks and splice sites.

Hugging Face

HViLM task models

Fine-tuned models for viral pathogenicity, transmissibility and host tropism.

Hugging Face

DNABERT-2

Efficient multi-species genome foundation model (117M parameters), co-developed with the MAGICS Lab.

Hugging Face
DatasetsProject page

HVUE benchmark

Benchmark for viral pathogenicity, transmissibility and host tropism prediction, with standard train and test splits.

Hugging Face
DatasetsProject page

GUE benchmark

Genome Understanding Evaluation: 36 datasets across 9 genome tasks, released with DNABERT-2.

GitHub
SoftwareProject page

deepvregulome

Python package to score how short variants affect the regulome, with motif-level explanations.

PyPI · GitHub
SoftwareProject page

ARGUS

Evidence-constrained agentic framework for interpreting noncoding variants.

GitHub
SoftwareProject page

DNABERT-2 code

Training and fine-tuning code for DNABERT-2.

GitHub
Web appsProject page

DeepVRegulome explorer

Interactive web app to explore variant effect predictions across cancer and neurodegeneration cohorts.

Streamlit

Using one of these in your work? Please cite the paper on its project page.