On a quantum-inspired kernel for classifying protein torsion angles

Abstract

Algorithms grounded in quantum principles need to demonstrate they are at least as expressive as established classical methods before any hardware advantage can be sought. Here we demonstrate that a quantum-inspired kernel reaches the same accuracy and Matthews correlation coefficient as carefully tuned radial basis function and degree-2 polynomial support vector machines when tasked with separating geometric regions in Ramachandran space. On a rigorously curated, balanced dataset of 10,000 torsion angle pairs derived from DSSP-annotated residues, the model achieves 98% accuracy with a Matthews correlation coefficient of 0.96, comparable to the top-performing classical models. By achieving true predictive parity on a well-characterised structural benchmark, the quantum-inspired kernel establishes that quantum-informed similarity measures already match classical baselines, laying a firm groundwork for future quantum-native bioinformatics workflows once suitable hardware becomes available.