AI & Protein Structures: New Benchmark for Accurate Predictions

February 18, 2026 Rachel Kim – Technology Editor Technology

A new artificial intelligence tool, D-I-TASSER, developed by researchers at the National University of Singapore, is demonstrating increased accuracy in predicting the three-dimensional structures of complex proteins, potentially accelerating breakthroughs in biomedical research and disease understanding. The system, which combines AI with physics simulations, achieved roughly 13 percent higher accuracy than existing leading methods in testing, according to a report released February 10, 2026.

Predicting protein structures is a longstanding challenge in biology. Protein function is directly tied to its shape, but experimentally determining these structures is a slow and resource-intensive process. AI-driven methods, particularly those leveraging deep learning and neural networks, have begun to revolutionize the field, as recognized by the 2024 Nobel Prizes in Chemistry and Physics, awarded for operate in this area.

D-I-TASSER addresses the complexity of multi-domain proteins – proteins composed of multiple independently folding structural units. The tool predicts the structures of individual protein segments before assembling a complete structural model. This approach allows for modeling across a significant portion of the human proteome, the entire set of proteins expressed by a human organism. Researchers, led by Zhang Yang, published their findings detailing the system’s capabilities.

The development builds on earlier advancements in AI-based protein structure prediction, notably the AlphaFold system. However, challenges remain in understanding the dynamics of protein folding and the behavior of non-globular proteins, including those involved in amyloid aggregation, a process linked to several neurodegenerative diseases. European initiatives, such as the COST Actions NGP-net and ML4NGP, are actively working to integrate AI with experimental data to address these gaps.

Future development of D-I-TASSER will focus on expanding its framework to model RNA structures and protein-protein interactions, including antibody-antigen complexes. The long-term goal, according to researchers, is to model the dynamic protein folding processes that occur inside living cells. The National University of Singapore team has not yet announced a timeline for the release of D-I-TASSER for broader scientific utilize.