Machine Learning for Toxicity Prediction Using Chemical Structures: Pillars for Success in the Real World DOI Creative Commons
Srijit Seal, Manas Mahale, Miguel García-Ortegón

et al.

Chemical Research in Toxicology, Journal Year: 2025, Volume and Issue: unknown

Published: May 2, 2025

Machine learning (ML) is increasingly valuable for predicting molecular properties and toxicity in drug discovery. However, toxicity-related end points have always been challenging to evaluate experimentally with respect vivo translation due the required resources human animal studies; this has impacted data availability field. ML can augment or even potentially replace traditional experimental processes depending on project phase specific goals of prediction. For instance, models be used select promising compounds on-target effects deselect those undesirable characteristics (e.g., off-target ineffective unfavorable pharmacokinetics). reliance not without risks, biases stemming from nonrepresentative training data, incompatible choice algorithm represent underlying poor model building validation approaches. This might lead inaccurate predictions, misinterpretation confidence ultimately suboptimal decision-making. Hence, understanding predictive validity utmost importance enable faster development timelines while improving quality decisions. perspective emphasizes need enhance application machine discovery, focusing well-defined sets prediction based small molecule structures. We focus five crucial pillars success ML-driven property prediction: (1) set selection, (2) structural representations, (3) algorithm, (4) validation, (5) predictions Understanding these key will foster collaboration coordination between researchers toxicologists, which help advance discovery development.

Language: Английский

Machine Learning for Toxicity Prediction Using Chemical Structures: Pillars for Success in the Real World DOI Creative Commons
Srijit Seal, Manas Mahale, Miguel García-Ortegón

et al.

Chemical Research in Toxicology, Journal Year: 2025, Volume and Issue: unknown

Published: May 2, 2025

Machine learning (ML) is increasingly valuable for predicting molecular properties and toxicity in drug discovery. However, toxicity-related end points have always been challenging to evaluate experimentally with respect vivo translation due the required resources human animal studies; this has impacted data availability field. ML can augment or even potentially replace traditional experimental processes depending on project phase specific goals of prediction. For instance, models be used select promising compounds on-target effects deselect those undesirable characteristics (e.g., off-target ineffective unfavorable pharmacokinetics). reliance not without risks, biases stemming from nonrepresentative training data, incompatible choice algorithm represent underlying poor model building validation approaches. This might lead inaccurate predictions, misinterpretation confidence ultimately suboptimal decision-making. Hence, understanding predictive validity utmost importance enable faster development timelines while improving quality decisions. perspective emphasizes need enhance application machine discovery, focusing well-defined sets prediction based small molecule structures. We focus five crucial pillars success ML-driven property prediction: (1) set selection, (2) structural representations, (3) algorithm, (4) validation, (5) predictions Understanding these key will foster collaboration coordination between researchers toxicologists, which help advance discovery development.

Language: Английский

Citations

0