Recognition of nucleophilic substitution reaction mechanisms of carboxylic esters based on support vector machine. (28th September 2016)

• Record Type: Journal Article
• Authors: Yuan, Hua; Chen, Chun‐Ni; Li, Meng‐Yang; Cao, Chen‐Zhong
• Is Part Of: Journal of physical organic chemistry. Volume 30:Number 7(2017)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2016-09-28
• Subjects: carboxylic esters -- nucleophilic substitution reaction -- reaction mechanism -- structural descriptors -- support vector machine (SVM); Chemistry, Physical organic -- Periodicals; 547.1
• DOI: 10.1002/poc.3658
• Languages: English
• ISSNs: 0894-3230

Abstract

Abstract: The nucleophilic substitution reactions of carboxylic esters (R 1 COOR 2 ) follow 2 possible mechanisms, namely, stepwise mechanism and concerted mechanism. The reaction mechanism is affected by the structures of the nonleaving group (NLG, ie, R 1 ), the leaving group (LG, ie, OR 2 ) of carboxylic esters, and the nucleophiles (Nu). The traditional approach for the determination of reaction mechanism by dynamic experiments is not only laborious and time consuming but also only applicable for the situation with single structural factor variation. Benefiting from the ample existing kinetic data, this article aimed to construct a molecular structure‐based classification model by support vector machine (SVM). The polarizability effect index of the NLG (PEI(NLG)), the p K a of the conjugate acid of the LG (p K a (LG)), and the p K a of the conjugate acid of the Nu (p K a (Nu)) were used to characterize the structural information of the 3 influencing factors, respectively. On the basis of these structural descriptors, the SVM classification model was established for the reaction mechanisms of 225 nucleophilic substitution reactions in the training set. The total prediction accuracy of 99.11% was achieved by 5‐fold cross validation. Then the reaction mechanisms of 56 reactions in test set A were predicted by this SVM model, and the prediction accuracy was up to 98.21%. Furthermore, test set B consisting of 24 reactions were used as another external data set for prediction, where the reaction mechanisms were controversially reported by different studies. The prediction results of test set B by the SVM model built in this article were amazingly consistent with the conclusions drawn by Um et al. Abstract : A support vector machine classification model for the nucleophilic substitution reaction mechanisms of carboxylic esters was built based on 3 structural descriptors including PEI(NLG), p K a (LG), and p K a (Nu), which are not only simple to be calculated but also of clear physical meaning. The total prediction accuracy of 99.11% was achieved for the training set and 98.21% for the test set. This study provided a theoretical support for the understanding and recognition of reaction mechanism of nucleophilic substitution based on the molecular structure.