Prediction and Chemical Interpretation of Singlet-Oxygen-Scavenging Activity of Small Molecule Compounds by Using Machine Learning

Abstract

A chemically explainable machine learning model was constructed with a small dataset to quantitatively predict the singlet-oxygen-scavenging ability. In this model, ensemble learning based on decision trees resulted in high accuracy. For explanatory variables, molecular descriptors by computational chemistry and Morgan fingerprints were used for achieving high accuracy and simple prediction. The singlet-oxygen-scavenging mechanism was explained by the feature importance obtained from machine learning outputs. The results are consistent with conventional chemical knowledge. The use of machine learning and reduction in the number of measurements for screening high-antioxidant-capacity compounds can considerably improve prediction accuracy and efficiency.

Machine-learning-assisted molecular design of phenylnaphthylamine-type antioxidants

“…A random seed was set in these RF models; then, the feature importance of each of the 13 descriptors and substituent positions was extracted to represent their contribution to AA. 69 The results of these models could help us identify which substitute groups and which positions significantly influence antioxidant properties. All this work was performed in the scikit-learn framework.…”

Low-cost machine learning prediction of excited state properties of iridium-centered phosphors

“…The feature sets can be categorized into those based on substructure/fingerprints (i.e., Morgan, Dice), those based on graph descriptors (i.e., whole-complex revised autocorrelations, referred to as RACs, 57 ligandonly RACs, and Coulomb-decay RACs, 58 referred to as CD-RACs), and those based on electronic structure calculations (i.e., xTB, PBEh, and B3LYP). For the substructure feature sets, we generated Morgan fingerprints, 59,60 which have been used previously in machine learning chemistry applications, [61][62][63][64][65][66] by one-hot encoding of groups of atoms in a structure. We computed these with a radius of three and 2,048 bits on the isolated CN and NN ligands to capture the presence and absence of chemical substructures.…”

“…, xTB, ωPBEh, and B3LYP). For the substructure feature sets, we generated Morgan fingerprints, 59,60 which have been used previously in machine learning chemistry applications, 61–66 by one-hot encoding of groups of atoms in a structure. We computed these with a radius of three and 2048 bits on the isolated CN and NN ligands to capture the presence and absence of chemical substructures.…”

Feature Selection for the Interpretation of Antioxidant Mechanisms in Plant Phenolics

“…This is typically determined as a value relative to different reference compounds in each assay. The reaction mechanisms and targets vary depending on the compound; therefore, many individual studies have examined the antioxidant capacities of compounds that have similar structures and mechanisms of scavenging ROS [ 3 , 4 ]. However, it is difficult to comprehensively analyze the structure–activity relationship (SAR) of compounds.…”

“…We used five different assays for antioxidant capacity data: oxygen radical absorbance capacity (ORAC) [ 3 ], singlet oxygen absorption capacity (SOAC) [ 4 ], 3-(4,5-dimethylthiazole-2-yl)-2,5-diphenyltetrazolium bromide assay (MTT) [ 20 ], 2,2’-azinobis(3-ethylbenzothiazoline-6-sulfonic acid) radical cation decolorization assay (ABTS) [ 21 ], and DPPH radical-scavenging capacity [ 12 ]. Antioxidant capacity data for compounds from five different test methods were used for machine learning.…”