AI-Augmented HPC is redefining structural biology
Organizations increasingly leverage AI to predict protein structures and accelerate drug discovery, and the demand for high-performance data platforms capable of managing massive datasets and complex model training workflows continues to grow.
OpenFold and AlphaFold are two of the most used applications in the AI drug discovery space. OpenFold was launched in 2021 as a PyTorch reimplementation of AlphaFold, allowing researchers the ability to train on their own data sets and modify code. Since then, the project has greatly expanded, with pretrained models, scalability improvements for training and inference, and the recent release of OpenFold3. The application is now used by academic and commercial organizations worldwide.
DDN has joined the OpenFold Consortium
DDN has a long history of providing the high-performance infrastructure supporting life sciences research globally. Customers like Roche, use DDN to decrease analysis turnaround time from 15 to 2 days. DDN is now a supporting member of the OpenFold Consortium, and will collaborate with leading researchers, technology partners, and life sciences organizations to help streamline access to advanced protein-folding technologies while reducing infrastructure complexity for scientific teams.
The OpenFold Consortium was founded in 2025 and has over 40 members (and growing) globally. The Consortium is dedicated to advancing open source tools for AI structure prediction, primary though the development and release of OpenFold. However, the project continues to evolve, and has now expanded from protein folding into additional areas or research, including protein-ligand, protein-nucleic acid, and broader biomolecular structure prediction.
DDN Infinia solves an unmet need for AI-driven protein structure prediction
OpenFold, like many AI-based applications, can rely heavily on Object Storage. DDN Infinia improves upon the traditional object store model by:
- Eliminating the small file penalty of Protein databases
- Enabling microsecond key lookups
- Providing optimal write throughput for checkpointing.
Infinia can present as both Object store and POSIX compatible filesystem, so any workload can be supported on a single system. Infinia becomes an active part of the pipeline instead of slow, hard to manage data repositories. The platform’s ability to deliver consistent performance from data ingestion through AI inference helps researchers move from raw data to scientific insight faster.
Meet DDN at Supercomputing
Heading to Supercomputing? Join us at DDN Convergence or visit us at booth #4211 to see how DDN helps life sciences organizations accelerate discovery.
Can’t make it to the show? Contact us to learn how DDN helps life sciences organizations accelerate discovery.