Premium
Predicting experimental electrophilicities from quantum and topological descriptors: A machine learning approach
Author(s) -
Hoffmann Guillaume,
Balcilar Muhammet,
Tognetti Vincent,
Héroux Pierre,
Gaüzère Benoît,
Adam Sébastien,
Joubert Laurent
Publication year - 2020
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.26376
Subject(s) - computer science , support vector machine , decision tree , density functional theory , machine learning , set (abstract data type) , graph theory , quantum chemical , artificial intelligence , quantum , graph , molecule , topology (electrical circuits) , theoretical computer science , computational chemistry , mathematics , chemistry , quantum mechanics , physics , combinatorics , programming language
In this paper, we assess the ability of various machine learning methods, either linear or non‐linear, to efficiently predict Mayr's experimental scale for electrophilicity. To this aim, molecular and atomic descriptors rooted in conceptual density functional theory and in the quantum theory of atoms‐in‐molecules as well as topological features defined within graph theory were evaluated for a large set of molecules widely used in organic chemistry. State‐of‐the‐art regression tools belonging to the support vector machines family and decision tree models were in particular considered and implemented. They afforded a promising predictive model, validating the use of such methodologies for the study of chemical reactivity.