Preprint—not peer reviewed
Physics-Based Molecular Fingerprints from Spectral Graph Theory Provide Efficient Geometry-Aware Measures of Chemical Similarity
These fingerprints build a complete molecular graph with weighted edges, then use Laplacian eigenvalues to encode three-dimensional structure. With community detection, the authors observe strong performance against representative baselines across datasets spanning organic, inorganic, biological, reticular, and reaction chemistry. It gives large-scale similarity work a fixed-length, interpretable way to retain information from molecular geometry.
AI–chemistryAI–materials Materials representationMolecular representation