Back

A foundational in vivo platform for predicting human health outcomes

Roseberry, T.; Krausz, T.; Williams, G.; Tingley, D.

2026-08-09 pharmacology and toxicology
10.64898/2026.08.03.742611 bioRxiv
Show abstract

Rodents remain the workhorse of preclinical drug development, yet often fail to predict human clinical outcomes. Existing alternatives are similarly constrained. Cells in culture cannot recapitulate whole-organism physiology, and larger mammals cannot be studied at comparable throughput. Here we present a scalable, information-dense platform that can predict a drugs long-term human clinical outcomes from 24 hours of rodent behavior. A novel home-cage system continuously records behavior, generating thousands of features per hour. Models are trained on human clinical trial data to map these features onto outcomes including gastrointestinal adverse events, cardiac toxicity, neuropsychiatric side effects, and long-term weight loss. In addition to being an order of magnitude faster, the platform provides more accurate clinical predictions than standard long-term preclinical experiments. The approach readily extends to other outcomes, enabling rodents to serve as quantitative models for human clinical prediction.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.