News

CASUS is part of the successful initiative called PROFOUND

The Helmholtz Association, Germany’s largest research organization, has announced the winners of its second Helmholtz Foundation Model Initiative (HFMI) call. The Helmholtz-Zentrum Dresden Rossendorf (HZDR) with its complex system research-focused institute CASUS is part of a successful consortium that will now build up a new foundation model targeted at the analysis of protein structures. More specifically, the winning proposal PROFOUND aims to revolutionize the way how protein structures are predicted. If successful, the project would catalyze research in fields like medicine and biotechnology.

Foundation models are artificial intelligence (AI) models that have been trained on a lot of data. ChatGPT is an example for a foundation model focused on language processing. Other examples focus on images, videos or sound. Also science has discovered the huge potential of foundation models to answer open questions in a variety of research fields as they are significantly more powerful and flexible than conventional AI models. The Helmholtz Association is currently funding seven pilot projects in the areas of weather and climate, carbon dioxide cycles, radiology, plankton-bound carbon, photovoltaic materials, cell biology and protein design with the Helmholtz Foundation Model Initiative (HFMI). The aim of the three-year initiative is to develop fully functional models.

Additional information:

Dr. Artur Yakimovich

Young Investigator Group Leader

Hinge motions predicted for a certain part of the protein trypsinogen.

Protein design-focused PROFOUND is among the three proposals that a panel of experts has selected for funding in late October 2024. Proteins are building materials and chemical tools involved in biological processes. It is the way proteins interact and function that makes all living systems so complex. The ability to design proteins has enormous potential, for basic biology many applied areas and medicine. This year’s Nobel Prize in Chemistry was awarded to researchers who have pioneered this field. Today, it is possible to predict the three-dimensional structure of the protein that encodes a gene sequence using AI.

PROFOUND will revolutionize protein design by overcoming a major hurdle faced by current AI models such as AlphaFold: they are limited to static protein structures. In reality, proteins are like nanomachines that constantly change shape to perform their biological work. PROFOUND aims to capture these motions. The project team will leverage large-scale molecular dynamics data to create an AI model that can predict these dynamic behaviors. This approach will allow the design proteins that not only perform specific tasks but also adapt over time. Envisioned breakthroughs like dynamic enzymes and programmable molecular machines could lead to innovations in therapeutics, sustainable materials, and next-generation biotechnologies. The PROFOUND team is made up of experts from Forschungszentrum Jülich, Helmholtz Munich, Helmholtz-Zentrum Berlin, and the Center of Advanced Systems Understanding CASUS at Helmholtz-Zentrum Dresden-Rossendorf.

At CASUS, the Machine Learning for Infection and Disease group led by Artur Yakimovich will contribute to the development of a suitable diffusion model architecture. “One of the avenues we aim to explore is along the lines of our Conditional Variational Diffusion Model, an AI model that improves the quality of images by reconstructing them from randomness,” says Yakimovich. “While we have shown in that project that the noise schedule hyperparameter can be learned from the data, the data basis within PROFOUND is different and other avenues we have in mind might prove more suitable. In any case, we are proud to be able to contribute to this timely, ambitious and promising project.”