rtemis/tcga-pancan-hiseq

TCGA PANCAN HiSeq gene expression

RNA-Seq gene expression levels of 20,531 genes in 801 tumor samples from The Cancer Genome Atlas (TCGA) Pan-Cancer (PANCAN) project, measured on the Illumina HiSeq platform. The samples cover five cancer types. The dataset is UCI Machine Learning Repository dataset 401 in a single Parquet file with one row per sample.

Structure

data/tcga-pancan-hiseq.parquet: 801 rows, 20,533 columns.

Column	Type	Description
sample_id	string	Sample identifier, sample_0 to sample_800.
gene_0 … gene_20530	double	RNA-Seq expression level of one gene.
cancer_type	categorical	Cancer type, as a TCGA study abbreviation.

cancer_type	Cancer type	Samples
BRCA	Breast invasive carcinoma	300
COAD	Colon adenocarcinoma	78
KIRC	Kidney renal clear cell carcinoma	146
LUAD	Lung adenocarcinoma	141
PRAD	Prostate adenocarcinoma	136

The dataset has one train split that holds all 801 samples.

Load

R - using rtemis (1.4.3 or later):

dat <- read("hf://datasets/rtemis/tcga-pancan-hiseq")

R - using nanoparquet

url <- "https://huggingface.co/datasets/rtemis/tcga-pancan-hiseq/resolve/main/data/tcga-pancan-hiseq.parquet"
path <- file.path(tempdir(), "tcga-pancan-hiseq.parquet")
options(timeout = 600)
download.file(url, path, mode = "wb")
dat <- nanoparquet::read_parquet(path)

Python - using polars:

import polars as pl
dat = pl.read_parquet("hf://datasets/rtemis/tcga-pancan-hiseq/data/tcga-pancan-hiseq.parquet")

Notes

• Genes carry the dummy names of the UCI dataset (gene_0 to gene_20530), in the order of the original Synapse submission (syn4301332), which lists the gene identifiers.
• 267 genes are zero in every sample.
• Expression values range from 0 to 20.78. The UCI documentation does not specify their normalization.
• The dataset contains no demographic or clinical variables.

Source and processing

The UCI files data.csv and labels.csv were combined column-wise after verifying that both list the same sample identifiers in the same order. The unnamed identifier column was named sample_id, and the label column Class was named cancer_type and stored as a categorical variable. Expression values are unchanged.

Croissant 1.1 metadata: croissant.json.

License

CC BY 4.0, the license of the UCI source dataset.

Citation

Cite the UCI dataset and the TCGA Pan-Cancer paper:

@misc{fiorini2016gene,
  author = {Fiorini, Samuele},
  title = {{gene expression cancer RNA-Seq}},
  year = {2016},
  howpublished = {UCI Machine Learning Repository},
  doi = {10.24432/C5R88H}
}

@article{weinstein2013cancer,
  author = {{The Cancer Genome Atlas Research Network} and Weinstein, John N. and Collisson, Eric A. and Mills, Gordon B. and Shaw, Kenna R. Mills and Ozenberger, Brad A. and Ellrott, Kyle and Shmulevich, Ilya and Sander, Chris and Stuart, Joshua M.},
  title = {The Cancer Genome Atlas Pan-Cancer analysis project},
  journal = {Nature Genetics},
  year = {2013},
  volume = {45},
  number = {10},
  pages = {1113--1120},
  doi = {10.1038/ng.2764}
}

Acknowledgment requested by the National Cancer Institute (citing TCGA data):

The results shown here are in whole or part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga.