Navigating Activity Cliffs with Neural Combinatorial Optimization: A Graph-Based Roadmap for Drug Discovery

In the drug discovery field, Graph Edit Distance (GED) has emerged as a powerful tool for calculating Activity Cliff Analysis (ACA). ACA is a crucial tool in drug discovery that helps identify pairs of compounds with similar structure while demonstrating vastly different biological activities. However, traditional GED algorithms face limitations in handling large and intricate molecular graphs. This paper demonstrates a novel approach that utilizes Neural Combinatorial Optimization (NCO) to optimize the GED algorithm, enabling it to efficiently calculate GED values, enhancing its performance. This proposed NCO-based GED algorithm is applied to activity cliff analysis, where it successfully identifies pairs of compounds with similar structures but different biological activities. This NCO framework offers a promising avenue for optimizing various computational algorithms in drug discovery and other domains. By successfully implementing this NCO-based approach, ACA and its impact on drug discovery could be revolutionised leading to faster, more accurate activity cliff identification, and ultimately, the potential for the creation of life-saving drugs.