Development of Multi-Task QSTR Models for Acute Toxicity Prediction Towards Daphnia magna Using Machine Learning in the OCHEM Platform

Abstract

This research employed a multi-task modeling approach to assess the acute toxicity of various chemicals through quantitative structure-toxicity relationship (QSTR) models. An expert system was constructed using several machine-learning techniques and was developed with resources from the publicly available Online Chemical Database and Modeling Environment (OCHEM). The study details the underlying assumptions and methodologies for model selection, descriptor identification, and the strategic development that contributed to the research's successful outcomes. The dataset utilized for QSTR modeling comprised 2678 compounds, with acute toxicity evaluations conducted on Daphnia magna organisms. The predictive performance of the QSTR models was validated through both cross-validation and external test sets. The consensus regression model shows strong predictive accuracy, with a coefficient of determination (q²) ranging from 0.74 to 0.77. The consensus prediction for the external evaluation set afforded high predictive power, achieving a q² value between 0.79 and 0.81. Furthermore, additional validation was achieved using experimental data from 20 compounds, showcasing robust predictive capabilities. Importantly, a considerable proportion of the toxicity values predicted by the models were in close agreement with results from in vivo studies, highlighting the reliability of the approach used.

Received: 11 December 2024 | Revised: 24 January 2025 | Accepted: 12 February 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The developed QSTR models are openly available in OCHEM at http://ochem.eu/article/164296. The data that support this work are available upon reasonable request to the corresponding author.

Author Contribution Statement

Vasyl Kovalishyn: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing. Diana Hodyna: Formal analysis, Investigation, Resources, Data curation, Writing – review & editing, Visualization. Larysa Metelytsia: Investigation, Resources, Writing – review & editing, Visualization, Supervision, Project administration.