DeepCyte launched the DeeTox Atlas™, a single-cell metabolomic reference atlas of drug toxicity mechanisms. Enterprise pilot programs with global pharmaceutical companies are expected to begin in the coming months.
DeeTox Atlas is described as a foundation dataset built from two independent single-cell metabolomics perturbation studies, spanning approximately 100 toxicant compounds, 300,000 cells, and ~500 metabolites per cell—over 3,000 single-cell measurements per compound across six biological replicates. Each compound is mapped to a curated four-level hierarchy of toxicity mechanisms anchored to established Adverse Outcome Pathways (AOPs), according to the company.
Rather than generating new wet-lab data for every compound, DeepCyte trains a foundation model capable of predicting toxicity mechanisms for compounds it has never measured, said Theodore Alexandrov, co-founder and CEO of DeepCyte, adding that as the atlas expands, the platform becomes increasingly scalable, reducing laboratory dependence while improving predictive performance.
“AI in toxicology is only as good as the biological data it learns from. DeeTox Atlas lets us find subtle molecular patterns tied to key toxicity mechanisms—patterns our validation studies show are expressed in small subpopulations of cells and are effectively invisible to methods lacking single-cell resolution—and turn them into predictive models,” explained Alexandrov. “Our vision is to move toxicology from reactive laboratory testing toward predictive, mechanism-based AI that surfaces and explains safety liabilities earlier in drug discovery.”
The Atlas will expand with more compounds, mechanisms, and biochemical and clinical data, making predictions more actionable for toxicologists, medicinal chemists, and safety scientists, noted Alexandrov.
