23.10.2025

DFG: Priority Programme “Artificial Intelligence for Protein Function (AI4PF)”

Deadline: 3 March 2026

Computational methods, particularly those based on Artificial Intelligence (AI), have recently led to transformative advances in protein science. The ability to predict native protein structures from amino-acid sequences has resolved long-standing challenges in computational biology. However, the protein-folding problem encompasses more than static structure prediction. Sequences encode dynamic folding pathways, environmental responsiveness, intracellular localisation and complex interactions with ligands including catalytic activity, conformational changes, for example, in transport mechanisms – functional dimensions that remain largely inaccessible to current AI-driven design approaches. The SPP 2546 AI4PF addresses these limitations by supporting research that develops new computational methods for the design of experimentally validated functional proteins by integrating AI with chemical and biophysical representations of the polypeptide chain.

For the purpose of this SPP, we define the following core areas: 

• development of new AI methods for functional protein design including customised functions and dynamic properties;
• expanding the designable sequence space beyond what nature has explored, e.g. by including non-canonical amino acids or posttranslational modifications;
• integration of AI methods with physical models to improve predictive power and design capabilities;
• surmounting limitations of restricted and heterogeneous training datasets, e.g. by creating or curating novel experimental datasets for training and validation; and
• overcoming evolutionary constraints embedded in existing protein data and AI methods to design proteins truly de novo.

Proposals will be evaluated according to these research areas. Addressing multiple areas will be a plus. We explicitly include the design of smaller proteins/peptides, as long as they form a tertiary structure/structured core/folded domain.

Computational method development and experimental validation should go hand in hand. Tandems consisting of a computational and an experimental PI are encouraged to tightly link both aspects. All projects must integrate experimental validation of designed proteins in terms of structure, dynamics or function. The ideal project includes characterisation of structure, dynamics or biophysics in addition to biochemical function of the designed protein.

Thus, to maintain the focus on this SPP, projects are deemed not responsive if they:

• fail to integrate innovative AI method development with experimental validation demonstrating a molecular link between computational design and resulting structure, dynamics and function;
• engage in classic small molecule drug discovery or focus on engineering of other biomolecules such as RNA or DNA;
• rely primarily on manual/classical/rational protein engineering;
• rely primarily on experimental approaches such as directed evolution;
• rely primarily on existing/standard computational methods or pipelines;
• focus on non-molecular approaches such as systems biology.

A networking meeting, including short presentations of the planned projects followed by discussions, will be held on 24/25 November 2025 at the University of Leipzig. Participation is optional. Applicants who would like to attend in this network meeting are asked to send an e-mail to wirkstoffentwicklungspam prevention@uni-leipzig.de by 5 November 2025 for organisational purposes.

Please submit your proposal to the DFG by 3 March 2026.

Further information:
https://www.dfg.de/de/aktuelles/neuigkeiten-themen/info-wissenschaft/2025/ifw-25-82