In a recent Nature Communications study, researchers from the University of Melbourne analyzed all 214 million AlphaFold2-predicted protein structures to uncover new patterns in how proteins function, evolve and cause disease.

Using Ampere-based A1 Oracle Cloud instances and applying topological data analysis, they built a map of the “Protein Universe” — identifying functional domains, revealing evolutionary relationships and pinpointing structural regions where mutations can have profound effects. This has the potential to accelerate drug discovery, guide genetic research and deepen our understanding of life at the molecular level.

Read more: Nature Communications: The Topological Properties of the Protein Universe

Cryo-electron microscopy (Cryo-EM) produces vast amounts of high-resolution image data that must be processed through computational pipelines. This is increasingly augmented by AI for image classification, particle picking and 3D reconstruction.

Using cryo-EM, scientists are creating 3D images of ribosomes, proteins and viruses and then stitching together thousands of these images to study how structures within a molecule move and interact. At Nogales Lab, these studies were often halted when the data center reached thermal capacity which required time for servers to be cooled before studies could be continued.

Nogales Lab replaced legacy x86 servers with Ampere-powered systems, cutting power use by 60% and increasing compute capacity by 40%. This allowed their AI-driven Cryo-EM workflows to run faster, process more image datasets without thermal throttling and integrate machine learning steps more effectively into their structural biology research.

Read more: Nogales Lab Case Study

Mosquito genomics research at MSOE relies on AI-enabled bioinformatics pipelines to detect patterns in massive DNA datasets and model potential genetic modifications. These pipelines involve sequential processing stages that require high memory capacity and consistent CPU throughput.

Switching to an Ampere-powered System76 Thelio Astra workstation with 128 cores and 512 GB RAM delivered a 10x acceleration in parallel processing, cutting analysis time from months to weeks. This made it possible to train and run AI models on significantly more genomic samples, enabling faster gene drive research to help prevent the spread of malaria and dengue.

Read more: MSOE Case Study

The University of Miami’s Systems Drug Discovery Lab uses physics-based simulations alongside AI-driven prediction models to study protein–ligand binding and guide the design of targeted protein degraders.

With ALAFIA’s AIVAS Supercomputer, powered by a 192-core AmpereOne® processor, the lab cut simulation times from over 24 hours to just a few, achieving up to 10× faster results. This accelerated both molecular dynamics simulations and the downstream AI analytics used to interpret structural data, making it easier to run iterative experiments and refine predictive drug models.

Read more: University of Miami Case Study

AI is no longer just a tool for automation. It’s a driver of discovery in every scientific domain. From training protein structure predictors to running large-scale genomic classifiers and drug-binding models, AI workloads demand massive parallelism, efficiency, and predictable, sustained performance.

By pairing AI’s potential with Ampere’s sustainable performance, researchers can explore more hypotheses, process larger datasets and deliver breakthroughs faster — all while keeping scientific computing efficient and accessible.