In silico prediction models for thyroid peroxidase inhibitors and their application to synthetic flavors

Abstract

Systematic toxicity tests are often waived for the synthetic flavors as they are added in a very small amount in foods. However, their safety for some endpoints such as endocrine disruption should be concerned as they are likely to be active in low levels. In this case, structure-activity-relationship (SAR) models are good alternatives. In this study, therefore, binary, ternary, and quaternary prediction models were designed using simple or complex machine-learning methods. Overall, hard-voting classifiers outperformed other methods. The test scores for the best binary, ternary, and quaternary models were 0.6635, 0.5083, and 0.5217, respectively. Along with model development, some substructures including primary aromatic amine, (enol)ether, phenol, heterocyclic sulfur, and heterocyclic nitrogen, dominantly occurred in the most highly active compounds. The best predicting models were applied to synthetic flavors, and 22 agents appeared to have a strong inhibitory potential towards TPO activities.

Supplementary information: The online version contains supplementary material available at 10.1007/s10068-022-01041-y.

Keywords: Machine learning; Quantitative structure–activity relationship (QSAR); Synthetic flavor; Thyroid peroxidase inhibitor (TPO); Toxicity prediction.

Fig. 1

Cross-validation (CV) and test scores for each feature, learning method, and grouping method (binary, ternary, or quaternary). For each grouping method, the black bar marked above with an asterisk indicates the test scores of the best-performing model. Each label is shown in “model name_feature extraction method (or voting methods in voting classifier)” form. [RF: Random forest; SVM: Support vector machine; ANN: Artificial neural network; AdaB: Adaptive boosting; XGB: Extreme gradient boosting; PCA: Principal component analysis; LDA: Linear discriminant analysis]

Fig. 2

Confusion matrices for the best-performing models of each grouping. The color intensity of each cell represents the proportion of the number of compounds predicted to be a class with respect to the actual number of compounds in a class

Fig. 3

Chemical space analysis between the input data for model development and synthetic flavors (red dots: synthetic flavors; green dots: input data for model development)