OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Alphagenome

Alphagenome is a public, MIT-licensed PyTorch implementation of AlphaGenome, a genomic attention model. It processes encoded DNA sequences into embeddings and, when configured with output heads, can return genomic track, contact and splicing predictions. The repository includes examples for human and mouse organism indices.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Alphagenome — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you want a Python/PyTorch implementation to adapt or evaluate in a genomics research workflow, particularly for sequence embeddings or track, contact and splicing outputs.

Who it is for

  • Computational biology researchers exploring attention-based genomic models.
  • Python developers building DNA-sequence modeling or splicing research workflows.

Documented uses

  • Generate embeddings from encoded DNA sequences.
  • Configure output heads for genomic tracks, contact outputs and splicing predictions.
  • Explore model training with the included minimal architecture example.
  • Load official JAX weights through the documented conversion utilities.
Before you choose

Requirements & limits

The repository is public and MIT licensed. Using pretrained Hugging Face weights requires an HF_TOKEN; the documented official-JAX conversion utilities require separately installing alphagenome-research.

  • The README gives concrete examples for 8,192-position DNA batches and human or mouse organism indices.
  • Output heads are configured by organism and can include tracks, contacts and splicing outputs.
  • Pretrained-weight use has separate prerequisites: a Hugging Face token for Hugging Face weights, or alphagenome-research for official JAX conversion.
  • The README identifies released official code and a lab continuation as other routes to consider.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install the package and configure Hugging Face access

Prepare the repository for using pretrained weights from Hugging Face.

Before you start

  • A local copy of the repository
  • Python package installation with pip or uv
  • A Hugging Face API token for pretrained-weight access

Steps

  1. Install the project with the command below or the command below.

    Documented command
    pip install .
    Documented command
    uv pip install .

  2. Copy .env.template to .env.

  3. Add the Hugging Face token as HF_TOKEN in .env.

Expected outcome

The package is installed and the environment has the credential required by the README for pretrained Hugging Face weights.

Guide limitations

  • This setup does not cover running inference or training.
  • The Hugging Face token is specifically required for pretrained-weight access.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
lucidrains/alphagenome · Official homepage title · retained in its source language
How we review sourcesSuggest a correction

Related options

Google DeepMind alphagenome_research — The repository identifies this as the official code and documents it as a dependency for converting official JAX weights.

genomicsxai/alphagenome-pytorch — The repository README points to this project as the continuation at Anshul Kundaje’s lab.