Knowledge-Based Artificial Intelligence System for Drug Prioritization.

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Traditional drug discovery is time-consuming and expensive.
• Artificial intelligence (AI) is revolutionizing drug development.
• In silico methods offer a faster, cost-effective alternative.

Purpose of the Study:

• To present a novel computational framework, labyrinth, for in silico drug prioritization.
• To simulate human knowledge retrieval and inference for identifying potential drug candidates.
• To provide an R package for accessible use of the labyrinth framework.

Main Methods:

• Integrated clinical trial data, literature co-occurrences, drug-target interactions, and disease similarities.
• Developed a computational framework designed to mimic human reasoning.
• Validated the framework across 20 disease categories using TCGA cohorts.

Main Results:

• Achieved over 90% predictive accuracy across clinical trial phases.
• Demonstrated strong alignment with clinical practice in TCGA cohorts.
• Showcased robust ROC-AUC metrics, balancing predictive accuracy and interpretability.

Conclusions:

• The labyrinth framework effectively prioritizes drug candidates for various diseases.
• The study highlights the importance of aligning computational models with human reasoning.
• The labyrinth R package is available on GitHub for broader application.