NVIDIA AI Updates: September 1, 2026

1. NVIDIA Brings BioNeMo NIM Microservices to Claude Science for Protein Structure Prediction

NVIDIA detailed how its BioNeMo Agent Toolkit integrates with Anthropic’s Claude Science, letting AI agents orchestrate protein structure prediction workflows through NVIDIA NIM microservices. The workflow chains three microservices: MSA Search for generating multiple-sequence alignments, plus the OpenFold3 and Boltz-2 structure prediction models. NVIDIA reported that evolutionary alignment data is the load-bearing input for interface prediction, with interface confidence (iPTM) reaching 0.85 for OpenFold3 and 0.82 for Boltz-2 when MSA data is supplied, but collapsing to roughly 0.14 to 0.19 without it. The example validated complex prediction accuracy on the Seh1 protein and a predicted Mio-family partner across the two independent models. Source