Synthetic accessibility scoring and its application to the high-throughput design of energetic molecules

Abstract

Synthetic accessibility assessment is the pivotal link bridging the conceptual design and the practical synthesis of energetic molecules. As the chemical space for exploration expands, traditional evaluation methods based on empirical intuition and reaction energy barrier calculations can hardly meet the demands of high-throughput virtual screening. In recent years, synthetic accessibility scoring models have gained much attention for their ability to provide rapid and more accurate evaluation. This review systematically explores the current landscape of synthetic accessibility scoring models, categorizing them into two primary types: molecular structure-based models (SAscore, SYBA, GASA, DeepSA, and BR-SAScore) and synthetic route-based models (SCScore, RAscore, RetroGNN, CMPNN, and DFRscore). Furthermore, the exploration of the application of synthetic accessibility scoring in the high-throughput combinatorial design of energetic molecules is elaborated, and three key issues are highlighted: (1) limited applicability of existing scoring models to energetic molecules; (2) insufficient data for constructing specialized synthetic accessibility scoring models for energetic molecules; and (3) the subjectivity of expert scoring labels. To address these challenges, potential research approaches are outlined, such as constructing synthetic accessibility scoring datasets for typical energetic molecules using the analytic hierarchy process, thus enabling the development of more accurate and reliable scoring models tailored to this field.