Berkeley Lab Develops AQuaRef: Faster, More Accurate Protein Structure Mapping

Lawrence Berkeley National Laboratory (Berkeley Lab) researchers, along with an international team, have developed a new computer program named AQuaRef. The program provides a faster and more accurate method for determining protein structures, utilizing quantum-mechanical (QM) calculations and artificial intelligence (AI) to generate higher quality structural data at reduced computational costs.

This advancement allows for a more precise understanding of molecular structures, potentially revealing new information about protein function in both healthy and diseased states.

AQuaRef is integrated into Phenix, a software suite widely used by structural biologists globally for macromolecular structure modeling. According to Berkeley Lab researcher Nigel Moriarty, understanding protein structure is critical for insights into disease mechanisms in humans and energy production in plants, which can lead to more effective therapeutics and bioenergy solutions.

Overcoming Current Limitations

The program addresses limitations in current protein mapping methods, which primarily rely on experimental data (X-ray crystallography, cryo-EM) and existing theoretical data libraries. These methods have been limited in defining meaningful noncovalent interactions. AQuaRef's approach overcomes these limitations by applying quantum and AI techniques.

Development and Proven Performance

Developed through a nearly five-year collaboration between the Phenix team and Carnegie Mellon University researchers, AQuaRef uses machine learning tools to compute energy and forces for proteins, making quantum-level refinement practical.

In 71 tested experiments, AQuaRef consistently produced higher quality structural information with lower computational expense while maintaining or improving fit to experimental data.

Impact and Future Applications

The program successfully determined proton positions in DJ-1, a human protein associated with Parkinson's Disease, whose structure has been challenging to map. Researchers plan to broaden AQuaRef's scope to include more diverse structures, including those relevant for pharmaceutical drug design. Potential applications extend to improving understanding of photosynthesis for crop productivity and mapping plant proteins for biofuel production.

Collaborators and Funding

Collaborators included the University of Wrocław, Poland, the University of Florida, and Pending.AI, Australia. The National Institutes of Health and the Phenix Industrial Consortium provided funding for this work.