New Python package opens the recount3 RNA-seq data trove to machine learning workflows

A new Python package now gives researchers direct programmatic access to recount3, an online resource holding tens of thousands of uniformly processed RNA-seq samples from human and mouse drawn from major sequencing repositories such as the Sequence Read Archive. Until now, working with these datasets has largely required the R/Bioconductor ecosystem, but the growing prominence of Python in bioinformatics and machine learning has created demand for native tooling. The recount3 Python package, described in BMC Genomics, fills that gap with a robust application programming interface and a command-line interface for discovering, downloading and materializing recount3 resources.

The package automates the plumbing that typically slows down large-scale transcriptomic analyses. It handles uniform resource locator resolution, maintains persistent on-disk caching, and automatically parses downloaded data into analysis-ready structures, including Pandas DataFrames and BiocPy RangedSummarizedExperiment objects. The package handles URL resolution, persistent on-disk caching, and automatic parsing of data into analysis-ready structures, so Python users do not have to carry out these steps manually.

The recount3 resource provides uniformly processed RNA-seq samples, meaning the data are processed with the same pipeline across samples. The new package gives Python users native access to these recount3 resources, which have traditionally been centered in the R/Bioconductor ecosystem.

The developers position the tool as a bridge between massive public transcriptomic data and modern machine learning ecosystems. The developers state that the package lowers the barrier to entry for large-scale utilization of RNA-seq data in Python-based computational pipelines, bridging the gap between massive public transcriptomic data and modern machine learning ecosystems.

For computational biologists, the practical implications are straightforward. A laboratory can script retrieval across the recount3 resource, cache the results locally to avoid repeated downloads, and load the data into Pandas DataFrames or BiocPy RangedSummarizedExperiment objects for downstream analysis. The command-line interface additionally supports users who prefer shell-based workflows or need to integrate recount3 retrieval into automated data pipelines.

The authors state that the package drastically lowers the barrier to entry for large-scale utilization of RNA-seq data in Python-based computational pipelines. By abstracting URL resolution, caching and parsing behind a single interface, the package reduces the technical steps required to access recount3 resources from Python.

The article is published as open access in BMC Genomics by Alexander Alsalihi, Robert M. Flight and Hunter N. B. Moseley. The package is presented as software, with the authors emphasizing that it drastically lowers the barrier to entry for large-scale utilization of RNA-seq data in Python-based computational pipelines.

Subject of Research: A Python package for programmatic access to the uniformly processed recount3 RNA-seq resource

Article Title: The recount3 Python package for programmatic access to uniformly processed RNA-seq data

Article References: Alsalihi, A., Flight, R. M., & Moseley, H. N. B. (2026). The recount3 Python package for programmatic access to uniformly processed RNA-seq data. BMC Genomics. https://doi.org/10.1186/s12864-026-13499-9

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13499-9

Keywords: recount3, RNA-seq, Python, bioinformatics, machine learning, Sequence Read Archive, BiocPy, Pandas, open science, transcriptomics, software, package