Software powered by AI has demonstrated the possibility to forecast the structure of proteins with great accuracy

Sep, 2021 - By WMR

Predicting protein structure can help researchers better understand how proteins work and speed up treatment for diseases such as COVID-19 and cancer.

RoseTTAFold, a software tool was developed with the collaboration of the Integrated Bioimaging Division and Berkeley Labs’s Molecular Biophysics. To produce the tool the partnership was led by the University of Washington. The algorithm used in the software considers distances, patterns, and coordinates of amino acids instantaneously. The tool evaluates the interactions inside and between the structures as these data inputs come in, eventually assisting in the construction of a very detailed representation of a protein structure.

To map a protein structure the current practices typically rely on advanced experiments at synchrotrons. Even these complex procedures have their limitations of quality and data which are not always possible to completely analyze a protein at the atomic level. It is now possible to determine a protein structure from its gene sequence using advanced machine learning processes applied to a huge library of protein shapes.

A subteam of structural biologists led by Paul Adams, Lab Director for Biosciences at Berkeley Lab utilized a RoseTTAFold prediction to solve the structure of a new protein from experimental observations they had previously acquired to check the accuracy of the forecasts. Randy Read of Cambridge University, who works with Paul Adams on the Phenix software application for automated molecular structural analysis, put the new system to the test on crystallographic data that had previously proven difficult to fix using traditional techniques.

Accurate structural models can contribute to a better understanding of how the disease is caused by mutations in certain proteins. More structural biology investigations will be catalyzed by these prediction techniques, according to Paul Adams, in order to understand the fine chemical features of proteins that other forecast methods can't provide. All of this will contribute to the advancement of the fundamental understanding of biology, the development of new and improved medications, and the engineering of proteins for the bioeconomy's increasing demand.