Predicting clinical trial outcomes using drug bioactivities through graph database integration and machine learning.

Area of Science:

• Pharmacology and Cheminformatics
• Computational Biology and Bioinformatics

Background:

• Clinical trial success rates significantly impact pharmaceutical research costs and timelines.
• Stringent regulatory processes increase pressure on drug development.
• Accurate prediction of trial outcomes is crucial for optimizing research workflows.

Purpose of the Study:

• To develop a machine learning (ML) approach for predicting clinical trial outcomes.
• To investigate the utility of biological activities as predictors for trial success.
• To create an integrated dataset linking clinical trial information with protein targets.

Main Methods:

• Utilized a Random Forest classifier, an ensemble learning method.
• Incorporated features such as biological activities, physicochemical properties, target-related features, and NLP-based compound representations.
• Extracted drug-disease pairs and mapped corresponding protein targets from multiple data sources.

Main Results:

• Achieved an average accuracy of 93% and an F1 score of 0.96 for predicting trial success ('Pass' class).
• Demonstrated that biological activity is a statistically significant predictor of clinical trial outcomes.
• Ensemble learning models outperformed smaller, independently trained ML models.

Conclusions:

• The developed ML model reliably predicts clinical trial outcomes, particularly success.
• Biological activity is a key factor in determining clinical trial success.
• The open-source dataset and code facilitate further research in predicting drug development success.