The presentations collectively showcase the best-in-industry predictive power of VivoSim's proprietary AI-enabled NAMkind™ Liver and NAMkind™ GI platforms. In liver toxicology testing, across a set of compounds where animal models and traditional methods provide 50-65% sensitivity, VivoSim’s models provided >90% sensitivity at detecting true positives for liver toxicity. Whereas current methodologies including animal in vivo testing can result in >10% false positives (wrongly predicted to have liver tox when they do not, causing rejection of a viable drug candidate), VivoSim’s liver toxicology methods result in fewer than 5% false positives.

These powerful results help VivoSim clients avoid expensive mistakes in drug development, at a time when the industry is undergoing a shift in preclinical safety evaluation. "With a heavy push from the US FDA, the pharmaceutical industry is moving away from traditional animal testing toward high-fidelity, human-relevant alternatives," said Keith Murphy, VivoSim Executive Chairman. "As our newest data show, using AI-informed models in highly accurate tissues, we detect the problems that pharma typically misses. When a pharma has a miss, the cost to them is many lost years and $50M - $200M, so strategic-thinking pharma executives are taking notice of the availability of better tools."

The United States Food and Drug Administration (FDA) is actively encouraging human-relevant NAMs in place of animal studies, a push that is resulting in greater pharma demand for new technologies. After presenting its April 2025 Roadmap to Reducing Animal Testing in Preclinical Safety Studies, the FDA recently reported that it is meeting its goals for stimulating pharma uptake of these new models. VivoSim’s novel models offer perhaps the most predictive tool to make FDA’s vision a reality.

VivoSim’s superior results are achieved by a combination of the best biological model with the most representative primary human cell types, leveraging multiple endpoints as readouts for best prediction, and training AI prediction models with multiple endpoints that provide a rich data set, yielding high prediction accuracy. In addition to testing traditional oral pill small molecules, VivoSim is establishing competitive advantages across new biotech modalities such as antibodies, siRNA, and gene therapies. Adding to the data supporting its industry leadership in small molecules as outlined above, the company has done extensive testing of Antibody Drug Conjugates (ADCs), one of oncology’s fastest-growing modalities, with comparable accuracy results in terms of predicting the clinical profiles of ADCs for liver toxicity and diarrhea. Globally, hundreds of ADC candidates are now in active clinical development against over 50 molecular targets — 41 of them already in Phase III — and the pipeline remains overwhelmingly oncology-focused, led by HER2- and TROP2-directed programs in breast and lung cancer. Therapies still fail in human trials, most often on safety or efficacy. VivoSim’s mission is to close gaps in the ability to predict outcomes before a drug enters expensive clinical trials.

Overall, the findings show that the Company’s human-relevant models can predict liver and gastrointestinal toxicity — for both traditional small molecules and complex antibody-drug conjugates (ADCs) — with accuracy that tracks real clinical outcomes. The data are being presented at The 23rd International Congress of the European Society of Toxicology In Vitro (ESTIV 2026) in Maastricht, the Netherlands, which takes place from June 29 to July 2, 2026.

Key highlights of the presented data demonstrating VivoSim’s ability to achieve definitive translational signatures across complex organ systems:

High-Accuracy Liver Profiling: Benchmarked against a clinical small-molecule dataset of 92 compounds, the NAMkind™ Liver spheroid model achieved a predictive accuracy of 91%, with 90% sensitivity, 95% specificity, and 99% precision under repeat-dose conditions. The multi-endpoint profiling suite effectively minimized false negatives and successfully resolved intra-class toxicities, such as within thiazolidinediones (TZDs).

 

Mechanistic GI Insights: Using the multicellular human intestinal barrier model (NAMkind™ GI), VivoSim successfully integrated multiple endpoints including barrier function to resolve distinct mechanistic classes of tyrosine kinase inhibitor (TKI)-induced diarrhea.

 

Deconvolution of ADC Toxicity: Critically, both platforms demonstrated a unique capacity to evaluate advanced modalities by differentiating ADC risk based on structural properties. The GI model established that Trastuzumab-deruxtecan-mediated injury is exposure-dependent and payload-driven through detailed histological tracking. Meanwhile, the Liver model successfully differentiated hepatic risk based on linker stability and payload permeability when comparing Trastuzumab Emtansine and Trastuzumab Deruxtecan.