Structural comparison of homomolecular systems on surfaces using a fingerprint-based method

Abstract

This work presents an adaptation of the Smooth Overlap of Atomic Positions (SOAP) method to improve the efficiency of (dis)similarity quantification in homogeneous molecular (homomolecular) systems. SOAP, a fingerprint-based approach, is widely used to measure molecular similarity. We propose variants of SOAP kernels that leverage the structural architecture of homomolecular systems to minimize irrelevant comparisons of atomic environments. To evaluate its performance, we apply this adapted SOAP-based method to a synthetic dataset consisting of two identical tripeptides deposited on a copper surface, simulating different molecular states. The results demonstrate that the adapted method not only improves computational efficiency but also yields more meaningful clustering outcomes by better capturing the key structural differences between states. These findings suggest that the proposed method is well-suited for the study of homomolecular systems, particularly those involving surface interactions, and has the potential to enhance the use of diverse types of molecular modeling and analysis methods that rely on (dis)similarity measures.