In Silico Models to Predict the Perturbation of Molecular Initiating Events Related to Thyroid Hormone Homeostasis

Abstract

Disturbance of the thyroid hormone homeostasis has been associated with adverse health effects such as goiters and impaired mental development in humans and thyroid tumors in rats. In vitro and in silico methods for predicting the effects of small molecules on thyroid hormone homeostasis are currently being explored as alternatives to animal experiments, but are still in an early stage of development. The aim of this work was the development of a battery of in silico models for a set of targets involved in molecular initiating events of thyroid hormone homeostasis: deiodinases 1, 2, and 3, thyroid peroxidase (TPO), thyroid hormone receptor (TR), sodium/iodide symporter, thyrotropin-releasing hormone receptor, and thyroid-stimulating hormone receptor. The training data sets were compiled from the ToxCast database and related scientific literature. Classical statistical approaches as well as several machine learning methods (including random forest, support vector machine, and neural networks) were explored in combination with three data balancing techniques. The models were trained on molecular descriptors and fingerprints and evaluated on holdout data. Furthermore, multi-task neural networks combining several end points were investigated as a possible way to improve the performance of models for which the experimental data available for model training are limited. Classifiers for TPO and TR performed particularly well, with F1 scores of 0.83 and 0.81 on the holdout data set, respectively. Models for the other studied targets yielded F1 scores of up to 0.77. An in-depth analysis of the reliability of predictions was performed for the most relevant models. All data sets used in this work for model development and validation are available in the Supporting Information.

Thyroid Disease Prediction Using Selective Features and Machine Learning Techniques

“…Several thyroid disease detection and classification approaches have been presented in the literature. For example, Garcia et al [ 9 ] predicted the high probable molecules initiating the thyroid hormone homeostasis using machine learning algorithms RF, LR, GBM, SVM, and deep neural networks (DNN). The early prediction of the molecules is helpful for further testing in the first stages of thyroid disease.…”

Quantitative Structure-Activity Relationship Modeling of the Amplex Ultrared Assay to Predict Thyroperoxidase Inhibitory Activity

“…The study reported an external BA equal to 0.85 that is analogous to the accuracy values observed on the TeS for the BRF and KNN models presented here, but the performance of the PLR model is associated a much lower prediction coverage (i.e., 54.5% for the PLR model). Recently Garcia de Lomana et al (2020) , proposed a set of models predicting the interference of small molecules with nine targets involved in the thyroid hormone homeostasis (deiodinase 1, 2 and 3, transtiretrin, TPO, thyroid releasing hormone receptor and thyroid stimulating hormone receptor). Models were developed using five machine learning algorithms in combination with three data balancing approaches.…”

In Silico Identification of Potential Thyroid Hormone System Disruptors among Chemicals in Human Serum and Chemicals with a High Exposure Index

“…(Quantitative) structure–activity relationship models [(Q)­SARs] are in silico models allowing screening of large chemical inventories based on quantitative data (QSAR) or qualitative data (SAR) where the potency of chemicals to induce an MIE are related to their structural and physico-chemical properties. Attempts have earlier been made to develop first-tier predictive tools to identify chemicals of concern for their potency to modulate the endocrine system, either focusing on one or several compound classes or targeting a limited number of protein targets related to different hormone systems. − Recently, a battery of QSAR models for some of the thyroid-specific MIEs was developed with several classical machine learning methods …”

“…21−23 Recently, a battery of QSAR models for some of the thyroid-specific MIEs was developed with several classical machine learning methods. 24 In the current study, our first aim was to develop first-tier predictive QSAR models able to identify potential THDCs with confidence using the Conformal Prediction (CP) framework with the underlying Random Forest. Models were built for all MIEs related to the TH system as recently described by Noyes et al, 11 where data for model development is available.…”