Proteins rarely function in isolation as individual monomers. Most biological processes are governed by proteins interacting with other proteins…

NVIDIA DevTalk serves as a vibrant community hub where developers can engage in discussions, seek assistance, and collaborate on projects involving NVIDIA hardware and software. Developers can tap into the collective expertise of the NVIDIA developer community, sharing insights, troubleshooting issues, and exploring best practices for GPU programming and AI development. Additionally, DevTalk provides a platform for developers to showcase their projects, receive feedback, and network with peers, fostering collaboration and knowledge exchange within the NVIDIA ecosystem.

NVIDIA Developer

NVIDIA researchers extended the AlphaFold Protein Structure Database (AFDB) with large-scale homomeric and heteromeric protein complex predictions using a high-throughput pipeline. The pipeline decouples MSA generation (via MMseqs2-GPU) from structure inference (via ColabFold and OpenFold accelerated with TensorRT and cuEquivariance), and orchestrates workloads across NVIDIA H100 DGX Superpod clusters using SLURM. Key optimizations include staggered colabfold_search processes to reduce GPU idle time, sequence packing to minimize JAX recompilations, and evidence-driven interaction selection via STRING. Accuracy validation on 125 PDB homodimers shows OpenFold with TensorRT/cuEquivariance matches ColabFold baseline quality (75.41% usable vs 72.95%, mean DockQ 0.647 vs 0.637). The resulting high-confidence complex structures are being made available through the AlphaFold Database.

How to Accelerate Protein Structure Prediction at Proteome-Scale