Texas A&M highlights AI tool for safer chemical screening

Researchers at Texas A&M’s College of Veterinary Medicine and Biomedical Sciences are spotlighting a new AI-based toxicology approach that doesn’t just predict whether a chemical may be harmful, but also estimates how confident the model is in that prediction. The work, highlighted by Texas A&M on May 27, 2026, builds on a January 7, 2026, Nature Communications paper led in part by VMBS professor Dr. Weihsueh Chiu. In that study, researchers developed uncertainty-aware machine learning models to predict non-cancer human toxicity for more than 100,000 chemicals on the global market, aiming to help identify which substances may warrant more testing or tighter regulatory scrutiny. (stories.tamu.edu)

Why it matters: For veterinary professionals, this is another example of how veterinary toxicology and comparative biomedical science are increasingly shaping broader public health decision-making. Texas A&M’s framing is especially relevant because the models are designed to support, not replace, expert review, helping regulators and scientists prioritize high-risk or high-uncertainty chemicals in a system where only a fraction of substances have been fully tested. That aligns with EPA’s continued use of predictive tools in chemical review and its stated push toward alternative methods that can reduce reliance on vertebrate animal testing. (stories.tamu.edu)

What to watch: The next question is whether uncertainty-aware AI models like this move from promising research into accepted regulatory workflows, where transparency, validation, and reporting standards remain major hurdles. (link.springer.com)