Modern insecticide substitutes using acetylcholine receptors (nAChR) as biochemical targets, such as neonicotinoid insecticides (NNIs), have been extensively researched. Only 12 compounds have been experimentally realized since the initial discovery of imidacloprid. Increasingly, the bottleneck in this field is to rapidly determine the synthesizability of NNI substitutes. Here, we designed a coupled evaluation system for synthesis prediction and validation, including the synthesis probability, reaction path difficulty, and electron transfer characteristics of NNIs and their substitutes. Firstly, a total of 1475 eigenvalues were generated and 52 critical eigenvalues were screened out through the Pearson's correlation coefficient. The positive and unlabeled (PU) machine learning was constructed using the critical eigenvalues NNIs, including 12 experimentally synthesized NNIs (positive samples) and 73 unsynthesized NNI substitutes (unlabeled samples). Results identified 3 NNI substitutes that were highly promising candidates for synthesis (synthesis probability > 0.5). The results of density functional theory demonstrated the ranking of their reaction ease was UN-1 (31.4 kcal/mol) > UN-2 (81.6 kcal/mol) > UN-3 (3.35 ×103 kcal/mol). Time-dependent density functional theory revealed that changes in the electron distribution and electron excitation type were critical factors affecting their synthesizability, and the local excitation type was more favorable for the synthesizability of NNI substituents. The findings provide significant guidance for NNIs synthesis, reducing the possible space of unlabeled samples to 95.89% of their original size, while also minimizing the cost of research on subsequent NNI substitutes.
Keywords: Density functional theory; Neonicotinoid insecticides substitute; Positive and unlabeled machine learning; Synthesis prediction; Time-dependent density functional theory.
Copyright © 2023 The Authors. Published by Elsevier Inc. All rights reserved.