NanoToxi AI

NANOTOXI AI HAS BEEN ACQUIRED BY INDO BIOACTIVE LABS (P) LTD. We built an AI-ML model that predicts nanoparticle cytotoxicity in under a second, accurately enough to skip early-stage in-vivo and in-vitro testing. We started research in July 2025 and in May 2026 our product became a part of Indo Bioactive Labs (P) Ltd.

We developed a novel AI-ML model and protocol for faster, biologically accurate cytotoxicity prediction of nanoparticles. It removes the early-stage need for in-vivo and in-vitro testing. A researcher submits a particle's profile (material, size, morphology, dose, exposure time, cell line, etc.) and gets back a classification, a confidence score, and a risk level in under 0.15 seconds. A four-model ensemble drives each prediction, grounded in real toxicological outcomes rather than category labels.

The point was triage. Rule candidate materials in or out before spending weeks of bench time and wet-lab budget on a molecule that was never going to work.

We led seven pilots with research institutions across India, Germany, and Nigeria, including the National Center for Nanoscience and Nanotechnology. Together they saved weeks of time and labor and approx. $50,000 in wet-lab testing.

Our models were trained on a curated dataset of over 25,000 nanoparticles, unified from literature and wet-lab research. Most of our work was data curation. Traditional nano-QSAR models tend to be narrow, covering a single cell line or metal oxides only (1), and report accuracies around 0.80 to 0.92. (2, 3)

• 97.8% prediction accuracy.
• Under 0.15s prediction speed.
• 64% fewer false positives than the traditional nano-QSAR standard.
• 4-model ensemble.

Emergent Ventures, Wharton Venture Labs.

1. Nano-QSAR modeling for metallic and metal oxide nanoparticles: a review. Ecotoxicology and Environmental Safety (2022).
2. NanoToxRadar: A Multitarget Nano-QSAR Model for Predicting Cytotoxicity of Multicomponent Nanoparticles. ACS Nanoscience Au (2025). R²_test ≈ 0.877.
3. A Nano-QSTR model to predict nano-cytotoxicity using human lung cell data. Beilstein J. Nanotechnol. / PMC10201760.