A Novel Graph-based Approach for Determining Molecular Similarity

TL;DR

This paper introduces a quantum-optimized, noise-tolerant graph similarity measure for molecules, improving mutagenicity prediction accuracy by accommodating measurement errors.

Contribution

It presents a novel quadratic unconstrained binary optimization formulation for graph similarity that incorporates noise relaxation, applied to molecular mutagenicity prediction.

Findings

Relaxed similarity measure improves prediction accuracy

Quantum annealer efficiently solves the optimization problem

Noise accommodation enhances robustness of molecular similarity assessment

Abstract

In this paper, we tackle the problem of measuring similarity among graphs that represent real objects with noisy data. To account for noise, we relax the definition of similarity using the maximum weighted co-$k$-plex relaxation method, which allows dissimilarities among graphs up to a predetermined level. We then formulate the problem as a novel quadratic unconstrained binary optimization problem that can be solved by a quantum annealer. The context of our study is molecular similarity where the presence of noise might be due to regular errors in measuring molecular features. We develop a similarity measure and use it to predict the mutagenicity of a molecule. Our results indicate that the relaxed similarity measure, designed to accommodate the regular errors, yields a higher prediction accuracy than the measure that ignores the noise.