VivoSim’s 3D Liver Model Hits 91% Accuracy, Cuts False Positives Below 5%
VIVS•VivoSim’s AI-enabled 3D NAMkind Liver model achieved >90% sensitivity and <5% false positives versus 50-65% sensitivity and >10% false positives in animal tests, helping avoid $50M-$200M drug development losses. In a 92-compound benchmark it posted 91% accuracy, 95% specificity and 99% precision, with similar GI and ADC insights.
1. Presentation Highlights
VivoSim showcased AI-enabled 3D NAMkind Liver and GI models at the European toxicology conference, demonstrating over 90% sensitivity and under 5% false positives in liver toxicity tests. The presentations also included ADC toxicity deconvolution, highlighting differentiated risk based on linker stability and payload properties.
2. Benchmark Performance
On a 92-compound small-molecule dataset under repeat-dose conditions, the NAMkind Liver spheroid model achieved 91% predictive accuracy, 90% sensitivity, 95% specificity and 99% precision. The GI platform resolved mechanistic classes of tyrosine kinase inhibitor–induced diarrhea and differentiated ADC profiles by structural characteristics.
3. Implications for Drug Development
These results position VivoSim to help clients avoid $50M to $200M in late-stage development failures by identifying liver and gastrointestinal toxicities early. With regulatory bodies encouraging human-relevant NAMs over animal testing, the company stands to capture growing demand for high-fidelity preclinical safety services.
