A decade of machine learning-based predictive models for human pharmacokinetics: Advances and challenges

Area of Science:

• Pharmacokinetics
• Drug Discovery
• Computational Chemistry

Background:

• Traditional in vitro and in vivo methods for estimating human pharmacokinetics (PK) are complex and expensive for large-scale compound screening.
• Existing artificial intelligence (AI) and big data approaches offer qualitative and quantitative prediction of drug PK but face challenges due to algorithmic adaptation and data limitations.

Purpose of the Study:

• To provide an overview of machine learning (ML)-based quantitative structure-activity relationship (QSAR) models for PK assessment.
• To highlight available databases for obtaining PK-related data.

Main Methods:

• Review of ML-based QSAR models applied to PK parameter prediction.
• Identification and compilation of relevant databases for drug PK data.

Main Results:

• Machine learning (ML) offers a viable approach for predicting pharmacokinetic (PK) values.
• Various databases exist to support the development and application of ML-QSAR models for drug discovery.

Conclusions:

• ML-based QSAR models represent a powerful tool for predicting human PK parameters, overcoming limitations of traditional methods.
• Leveraging big data and AI accelerates the drug development process by enabling efficient compound assessment.