Unified human gene representations

Gene Embeddings

Browse standardized gene-level representations across sequence, expression, literature, pathway, network, regulatory, knowledge-graph, and single-cell modalities.

Contribute an embedding Open Core Dataset

Community contributions are automatically checked for compliance with the technical policy. Inclusion is not independent scientific or legal endorsement.

— embedding sets
— maximum genes
— dimension range

Community contribution

Submit a gene embedding

Prepare one standardized package and upload it directly to the library-owned Hugging Face Dataset as a Pull Request.

1

Prepare one folder

<embedding_id>/
├── embeddings.npz
├── genes.tsv
├── README.md
└── submission.yaml

The embedding ID must contain 3–80 lowercase letters, numbers, dots, underscores, or hyphens and must begin with a letter or number.

A metadata.json file is automatically generated and saved in the embedding's folder after acceptance.

2

Check the package contract

  • embeddings.npz contains exactly one two-dimensional float32 array named embeddings.
  • genes.tsv contains one unique ensembl_gene_id column whose rows match the matrix rows.
  • submission.yaml provides the scientific context and provenance needed to understand the embedding and how it was produced.
  • README.md documents how the representation was produced.

Open gene mapping tutorial

3

Complete the submission template

Download submission.yaml, replace every placeholder and keep the embedding ID identical to the folder name. Submitted values are checked automatically against the package.

4

Upload as a Pull Request

  1. Open the Hugging Face uploader using the button below.
  2. Upload the folder <embedding_id>.
  3. Select Open as a pull request.
5

Automatic validation and publication

The package structure, metadata consistency, Ensembl identifiers, matrix shape, numeric validity, licensing, provenance, and attestations are checked automatically.

A strict PASS is stored in the central Dataset and added to this catalog.

Open Hugging Face uploader Download submission template Download completed example
Command-line alternative example

Before running this example, replace esm2_v1_1 with the name of your existing local embedding folder in both path arguments. 

hf upload \
  embeddingsSSRF/gene_embeddings_staging \
  ./esm2_v1_1 \
  esm2_v1_1 \
  --repo-type dataset \
  --create-pr \
  --commit-message "Submit gene embedding"