Run NVIDIA BioNeMo NIM Microservices for Protein Structure Prediction in Claude Science
- Source
- developer.nvidia.com
- Author
- Michelle Horton
- Date
Context
NVIDIA's BioNeMo Toolkit gives AI agents instructions for operating specialist scientific models. Its protein-prediction tutorial uses Anthropic's Claude Science to retrieve two protein sequences, build alignments of related sequences from other organisms, and run two structure-prediction models, OpenFold3 and Boltz-2. The alignments supply evolutionary information that can help the models predict where proteins interact.
In the tutorial's fungal-protein example, the interface confidence score (iPTM) for the predicted contact between the two proteins was 0.85 for OpenFold3 and 0.82 for Boltz-2 with input, and fell to 0.14 and 0.19 without it. The workflow saves inputs and outputs so researchers can inspect that difference. NVIDIA presents this as a reproducible way to form and inspect a structural hypothesis; the proteins' actual association has not been verified experimentally.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- alignment — The work of making AI systems actually pursue what their builders and users intend, rather than something subtly or dangerously different.
Checking sign-in…
Loading comments…