LSFG is a novel method for engineering proteins by utilizing an atomistic grid-based computational method that analyzes and compares protein atomic compositions. By arranging a protein’s atoms into a fine 3D grid, this approach captures atomic-level details in highly localized regions, allowing comparison of specific protein regions even in the absence of structural similarity. Unlike conventional methods that require spatial alignment, this approach maps chemical properties directly onto the grid, enabling alignment-free comparisons based on chemical composition. The method involves constructing a high-resolution grid around a protein’s 3D structure, where each grid point captures potential energy across the protein and allows to identify high-energy residues and functional regions, which are then targeted for engineering. Localized spherical feature grids (LSFGs) centered around these regions store atomic properties, enabling comparison with a database of known protein grids. Two comparison techniques are applied: geometric alignment using rotation matrices and quaternions, and transformer-based similarity scoring. High-ranking matches guide functional and stability optimization through mutation design. Applications include enhancing protein stability and functionality, predicting protein function, and engineering enzymes like glucose dehydrogenase (GDH) for improved co-factor recycling in biocatalysis. This method provides alignment-free chemical profiling for structurally diverse proteins, facilitating advanced protein engineering and functional annotation.
Patent No.: 581824
Date of grant: 26-02-2026