pkCSM: Predicting Small-Molecule Pharmacokinetic and Toxicity Properties Using Graph-Based Signatures

Area of Science:

• Computational chemistry
• Drug discovery and development
• Pharmacokinetics and toxicology

Background:

• High attrition rates in drug development are often caused by poor pharmacokinetic (PK) and safety profiles.
• Minimizing these risks early in the pipeline is crucial for efficient drug discovery.

Purpose of the Study:

• To introduce pkCSM, a novel computational approach for predicting key ADMET properties.
• To provide an accessible web server for rapid evaluation of drug candidates' PK and toxicity.

Main Methods:

• Development of predictive models using graph-based signatures.
• Utilizing computational approaches to assess ADMET properties.

Main Results:

• pkCSM models demonstrate performance comparable to or exceeding existing methods.
• The freely accessible web server offers an integrated platform for property evaluation.

Conclusions:

• pkCSM offers a valuable tool for early-stage drug development by predicting critical ADMET properties.
• The web server facilitates rapid assessment, potentially reducing drug development attrition.