This repository contains the curation of the metadata for SRA studies from the recount3 project first presented in the paper PLACEHOLDER.
The curation, available as a gzipped CSV file in curation/SRA/human/bulk/whole_tissues/curation.csv.gz, includes samples in SRA studies coming from whole human tissues sequenced using bulk RNA-seq protocols: 28,750 samples across 466 studies, one row per sequencing run. The full protocol is described in the README.md file inside the directory.
The file is stored gzipped (1.6 MB, against 115 MB unpacked) and can be read without unpacking it, by any tool that reads gzipped CSVs:
import pandas as pd
df = pd.read_csv("curation/SRA/human/bulk/whole_tissues/curation.csv.gz",
index_col=0, low_memory=False)Or, to unpack it: gunzip -k curation/SRA/human/bulk/whole_tissues/curation.csv.gz.
The utility scripts and configuration files used to generate the curation are available in the scripts and configs directories, respectively.
configs/SRA_studies_all.yaml is where the curation is actually written down: per study, the rules that say which of its samples are kept and what each of their fields means. The CSV above is that config applied to the metadata recount3 holds for each study, and can be rebuilt from it with scripts/get_dataset.py — see scripts/README.md.
A self-contained Streamlit app for interactively browsing the curated
metadata (filtering by organ/tissue, condition, age, sex/gender, and
drilling into individual studies) is available in
webserver/.
Try it live: https://recount3-sra-explorer-242644878949.europe-north1.run.app/
To run it locally instead, see the webserver/ README.