Accelerating the Transformation of Highly Energetic Molecules from “on Paper” to Synthesizable Targets: Synthetic Accessibility Assessment with Scoring Models

Data-driven design strategies have enabled the rapid generation of numerous energetic candidates, making efficient screening for synthesizable candidates critically important. Synthetic accessibility scoring models are widely applied in drug discovery, yet their applicability to energetic materials has not been systematically validated. Here, we present the energetic compound data set 100 (ECD100) based on the synthetic status-based labeling approach, a reproducible benchmark data set comprising 50 experimentally synthesized (easy to synthesize, ES) and 50 designed but unrealized (hard to synthesize, HS) energetic molecules. Eight representative SA models were benchmarked, among which SAscore achieved the best classification performance (ROC-AUC = 0.76). Further analysis identified fragmental structure distributions and molecular complexity as key contributors to model refinement. Furthermore, Pareto front analysis uncovered a trade-off between synthetic accessibility and detonation velocity, highlighting the need for multiobjective screening strategies. This work conducted an application evaluation of the synthetic accessibility scoring model in the field of energetic materials and provided new ideas for high-throughput screening of highly energetic molecules.