Stanford Releases ENCODE GRAMMAR: Thousands of AI Models to Decode Human Genome

The Kundaje Lab at Stanford University, in collaboration with the NIH ENCODE consortium, has released ENCODE GRAMMAR, a massive repository of deep learning models designed to decode the DNA sequence logic of regulatory elements within the human genome. This multi-year project serves as a vital comprehensive data resource for genomic research.

Confirmed

Project Scale: ENCODE GRAMMAR (Encyclopedia of Genomic Regulatory sequence Models, Motifs, Annotations, and Rules) comprises 3,865 experiment-specific deep learning model sets, covering data from over 2,000 experiments.
Scientific Background: Nearly all human cells share the same genome yet differentiate into cells with diverse forms and functions. The research highlights that this diversity is driven by specific regulatory sequences bound by transcription factors, and this project aims to leverage AI to decode these genomic regulatory codes.
Resource Content: Beyond the model sets, the project provides comprehensive data, including corresponding sequence annotations, motifs, and rules across multiple levels of gene regulation.

Why It Matters

By deeply integrating deep learning with genomics, this project helps researchers understand complex gene regulatory networks, providing a systematic AI toolset and data atlas for exploring cell differentiation and disease mechanisms.

2026-08-05 ~ 2026-08-06 · 12 related posts

Primary sources

ENCODE Releases GRAMMAR: 3,865 Deep Learning Models for Decoding Human Regulatory DNA — anshulkundaje · 2026-08-05 10:58 (UTC+08:00)
Stanford Releases ENCODE GRAMMAR: 3,865 Deep Learning Models to Decode the Human Genome — anshulkundaje · 2026-08-05 10:58 (UTC+08:00)
ENCODE Releases 2,300+ AI Models for Gene Regulation on Hugging Face — anshulkundaje · 2026-08-05 11:29 (UTC+08:00)

ENCODE GRAMMAR: A Genomic Regulatory Atlas of Sequence Models and Rules Announced — anshulkundaje · 2026-08-05
[source] ENCODE Releases GRAMMAR: 3,865 Deep Learning Models for Decoding Human Regulatory DNA — anshulkundaje · 2026-08-05
Stanford Team Introduces GRAMMAR: Decoding Genomic Regulation with AI — anshulkundaje · 2026-08-05
ENCODE GRAMMAR: Nearly 4,000 AI Models for Regulatory DNA Sequences Released — anshulkundaje · 2026-08-05
[source] ENCODE Releases 2,300+ AI Models for Gene Regulation on Hugging Face — anshulkundaje · 2026-08-05
Introducing ENCODE GRAMMAR: 3,865 Models for Regulatory DNA — anshulkundaje · 2026-08-06
ENCODE Releases Atlas of 3,865 Deep Learning Models for Human Regulatory DNA — anshulkundaje · 2026-08-06

5 near-duplicate retellings: anshulkundaje · anshulkundaje · anshulkundaje · anshulkundaje · anshulkundaje