README.md

metadata

configs:
  - config_name: gene_info
    data_files:
      - split: train
        path: gene_info/train-*
  - config_name: generated_descriptions_gpt4o_mini_combined
    data_files:
      - split: train
        path: generated_descriptions_gpt4o_mini_combined/train-*
  - config_name: ncbi_summary
    data_files:
      - split: train
        path: ncbi_summary/train-*
  - config_name: ncbi_uniprot_summary
    data_files:
      - split: train
        path: ncbi_uniprot_summary/train-*
dataset_info:
  - config_name: gene_info
    features:
      - name: gene_id
        dtype: string
      - name: ensembl_id
        dtype: string
      - name: gene_type
        dtype: string
    splits:
      - name: train
        num_bytes: 3580096
        num_examples: 84425
    download_size: 1305766
    dataset_size: 3580096
  - config_name: generated_descriptions_gpt4o_mini_combined
    features:
      - name: generated_description
        dtype: string
      - name: gene_id
        dtype: string
    splits:
      - name: train
        num_bytes: 108207019
        num_examples: 33703
    download_size: 46780098
    dataset_size: 108207019
  - config_name: ncbi_summary
    features:
      - name: description
        dtype: string
      - name: gene_id
        dtype: string
    splits:
      - name: train
        num_bytes: 10982104
        num_examples: 33703
    download_size: 3737904
    dataset_size: 10982104
  - config_name: ncbi_uniprot_summary
    features:
      - name: description
        dtype: string
      - name: gene_id
        dtype: string
    splits:
      - name: train
        num_bytes: 19163326
        num_examples: 33703
    download_size: 7356592
    dataset_size: 19163326

---

Gene Description Dataset

This dataset reproduces and expands upon the GenePT project and paper, and makes it easier to reproce and access using standard tools. The goal is to allow users to compose embeddings across dimensions in order to specialize for specific tasks, and add to the existing base embeddings by generating new descriptions and embedding them in the same space.

Dataset Description

This dataset

1. reproduces the data from the GenePT project and paper in a more easy-to-use format. This paper itself aggregates data from multiple sources, so please refer to the paper and repository for detailed source information. Citation: Chen YT, Zou J. (2023+) GenePT: A Simple But Effective Foundation Model for Genes and Cells Built From ChatGPT. bioRxiv preprint: https://www.biorxiv.org/content/10.1101/2023.10.16.562533v2. GitHub: https://github.com/yiqunchen/GenePT
2. Adds descriptions of genes extracted from various LLMs, across multiple dimensions, such as regulatory pathways, drug interactions, etc. Currently we use GPT-4o-mini to generate descriptions, and only have a combined description that includes several factors. We will add composable dimensions soon

Dataset Structure

The dataset contains four main components:

1. NCBI Summary

• Contains gene descriptions from NCBI

2. NCBI-UniProt Summary

• Contains combined gene descriptions from NCBI and UniProt

3. Gene Info Table

• Contains contains a mapping between gene_id, ensmble_id and gene functional annotation

4. Generated Descriptions (Combined)

• Contains AI-generated gene descriptions
• Model: GPT-4o-mini
• Factors: a. Associated genes b. Aging related information c. Drug interactions d. Pathways and biological processes

Dataset Creation

Source Data:

• NCBI Gene Database
• UniProt Database