Federal Project Grant Award Summary

The National Science Foundation (NSF) Division of Biological Infrastructure awarded $631,195 to the University of Missouri System (Curators of the University of Missouri) under the Biological Sciences program (CFDA 47.074) on August 15, 2026, for research on deep comparative learning methods to estimate protein structural model accuracy. The project will develop an advanced artificial intelligence (AI)-based tool and open-source software to help researchers assess and rank predicted protein structures—addressing the critical challenge that scientists often generate multiple structural models for the same protein but lack reliable methods to determine which predictions are most accurate. This capability is essential for leveraging recent breakthroughs in AI-based protein structure prediction and enabling reliable protein-based scientific discovery and bioengineering applications.

Over the three-year performance period (through July 31, 2029), the project will deliver three primary research products: (1) a large-scale benchmark dataset of comprehensive protein structural models for training and evaluating machine learning methods; (2) deep comparative learning methods for assessing monomer protein model quality; and (3) extended methods applicable to multimer protein complexes. The research will incorporate state-of-the-art AI techniques including protein language models, graph transformers, multi-task learning, and novel pairwise ranking losses. Additionally, the project includes workforce development components targeting undergraduate students, K-12 participants, and community audiences to strengthen computational biology training and advance STEM participation across educational levels.