Topos Bio offers an AI‑native platform that uses the all‑atom Topos‑1 model to generate high‑throughput conformational ensembles of intrinsically disordered proteins, revealing transient binding pockets for drug design. The system couples generative chemistry to propose small‑molecule candidates and incorporates wet‑lab feedback to continuously refine predictions, enabling pharmaceutical teams to target previously undruggable proteins in neurodegenerative and cancer programs.
Funding
Funding not disclosed
Founders
Product
Problem
Traditional drug discovery relies on static protein structures to identify binding pockets, but intrinsically disordered proteins (IDPs) constantly change shape and lack stable pockets, rendering roughly one‑third of the human proteome—including key drivers of Alzheimer’s, Parkinson’s, and aggressive cancers—effectively undruggable.
Solution
Topos Bio has built an AI‑native platform centered on Topos‑1, an all‑atom foundation model that learns the distribution of conformations adopted by IDPs and generates realistic ensemble representations at research‑scale speed. By sampling millions of protein shapes, the model reveals transient binding opportunities that can be targeted with small‑molecule designs. The platform couples generative chemistry to propose candidate compounds and integrates wet‑lab validation loops that feed experimental data back into the model, continuously improving accuracy. This ensemble‑focused workflow enables rational drug design against dynamic targets, opening therapeutic avenues for neurodegenerative and oncologic diseases previously considered “undruggable.”
Target Audience
Primary customers are pharmaceutical and biotech research teams focused on neurodegenerative, oncology, and metabolic disease programs that target intrinsically disordered proteins.
Features
- All‑atom foundation model (Topos‑1) that generates physically realistic conformational ensembles for intrinsically disordered proteins
- High‑throughput ensemble generation fast enough for day‑to‑day research and drug‑discovery cycles
- Generative chemistry engine that designs small‑molecule candidates optimized for dynamic binding sites identified in ensembles
- Integrated wet‑lab feedback loop that validates predictions and refines the model with proprietary experimental and simulation data
- Proprietary dataset combining large‑scale simulations and experimental measurements specific to IDPs, enabling training on a class of proteins lacking public data
- Benchmarking against leading protein models (e.g., AlphaFold, Chai) showing superior performance on disordered protein prediction