Developing Robust Human Liver Microsomal Stability Prediction Models: Leveraging Inter-Species Correlation with Rat...

Area of Science:

• Drug Discovery and Development
• Pharmacokinetics and Drug Metabolism
• Computational Chemistry and Cheminformatics

Background:

• Pharmacokinetic issues are a major cause of drug attrition, necessitating early-stage evaluation.
• In vitro assays like microsomal stability are crucial for assessing drug pharmacokinetic profiles.
• High-throughput screening generates extensive data, but resource-intensive assays like human liver microsomal (HLM) stability demand alternative strategies.

Purpose of the Study:

• To develop a predictive in silico model for human liver microsomal (HLM) stability.
• To leverage large-scale experimental data for training machine learning models.
• To provide an accessible and accurate tool for early-stage drug discovery.

Main Methods:

• Development of a HLM stability prediction model using an in-house dataset.
• Application of classical machine learning and advanced techniques, including neural networks.
• Validation of the model using external test sets and comparison with existing literature models.

Main Results:

• Achieved model accuracies exceeding 80% for HLM stability prediction.
• Demonstrated comparable performance to state-of-the-art models in the literature.
• Observed improved HLM model performance when incorporating rat liver microsomal (RLM) stability predictions, highlighting cross-species data utility.

Conclusions:

• The developed HLM stability prediction model offers a valuable tool for the drug discovery community.
• The model and a subset of the dataset are publicly accessible via the ADME@NCATS website.
• This represents the largest open-source model of its kind, uniquely leveraging cross-species data.