uc-ctds/GDC-QAG-genes-mutations

Dataset Card for uc-ctds/GDC-QAG-genes-mutations

This dataset contains genes and somatic mutations observed in various cancers. It is scraped from the /ssms endpoint in the Genomic Data Commons (GDC). This data is used to run the Query Augmented Generation (GDC) tool on the GDC. **GDC QAG** is currently deployed in the HuggingFace Spaces as a web app.

Dataset Details

Dataset Description

This dataset contains around 5.6 million somatic mutations (protein variants) observed in n=19606 genes in various cancers. The /ssms endpoint is queried using fields=gene_aa_change to obtain genes and protein variants observed in the various GDC projects. Please refer to our GitHub repo for details on dataset curation. All data is open-source.

• —Curated by: Center for Translational Data Science
• —Language(s) (NLP): en
• —License: apache-2.0

Dataset Sources [optional]

• —Repository: https://huggingface.co/datasets/uc-ctds/GDC-QAG-genes-mutations
• —Paper: coming up
• —Demo: **GDC QAG**

Uses

Direct Use

This dataset is meant to be used by the QAG tool. A dictionary is pre-constructed from this dataset and used at inference time to map a natural language query into gene and mutation entities.

Dataset Structure

The dataset is organized in this format:

{
    "BRPF3": [
        "D138=",
        "D165=",
        "F675V",
        "T255S",
        "T282S",
        "S1198=",
        "S864=",
        ....
      ],
    "GAS7": [
        "Q238*",
        "Q314*",
        "Q318*",
        "Q378*",
        "P17R",
        "P21R",
        ...
      ]
}

Dataset Creation

Curation Rationale

This dataset was curated to enable a fast mapping of genes and mutations mentioned in a natural language query to that observed in the GDC.

Source Data

The Genomic Data Commons API (/ssms endpoint)

Personal and Sensitive Information

None

Citation

Coming up soon

BibTeX: