PUBHOMICS: A Multispecies Biological Dataset to Catalyze AI-Driven Toxicity Assessment for Environmental and Public Health

Daniel Chinwendu Ukaegbu

Track: Track 2: Dataset Proposal Competition

Keywords: Transcriptomics, ML, AI, Chemical safety, and Toxicology, Public and Enviromental health.

TL;DR: PUBHOMICS is a large-scale multispecies dataset designed to accelerate AI-driven toxicity assessment for public and environmental health.

Abstract:

Environmental and public health remain under served by the recent data revolution that enabled
major AI advances in drug discovery. Existing toxicity datasets are biased toward drug-like molecules and
are fragmented across repositories, limiting their use for machine learning and cross-species translation.
We propose PUBHOMICS, a scalable, openly shareable dataset capturing transcriptional responses to
environmentally relevant chemical perturbations across cell types, organs, and species. PUBHOMICS
will expand chemical coverage to classes absent from existing resources, enable AI models to predict
transcriptomic responses to novel exposures, and support mechanism-based toxicity prediction with
cross-species translation for regulatory decision-making. By advancing exposomics toward causation and
providing a foundation for New Approach Methodologies (NAMs), PUBHOMICS aims to accelerate
regulatory adoption and enable “benign-by-design” strategies that bridge exposure science with systems
biology.

Submission Number: 185