Predicting drug adverse effects using a new Gastro-Intestinal Pacemaker Activity Drug Database (GIPADD)

Area of Science:

• Pharmacology
• Computational Biology
• Biophysics

Background:

• Artificial intelligence (AI) and big data offer new avenues for drug discovery and safety assessment.
• Understanding drug effects on gastrointestinal (GI) electrical activity is crucial for predicting adverse effects (AEs).
• Existing databases may not fully capture the complex electrical profiles of drug interactions.

Purpose of the Study:

• To evaluate the potential of the Gastro-Intestinal Pacemaker Activity Drug Database (GIPADD) for predicting drug AEs using machine learning (ML).
• To explore correlations between GI electrical features (EFs) and drug-induced AEs.
• To establish a novel method for clustering drugs based on their electrical activity profiles.

Main Methods:

• A comprehensive database (GIPADD) was created using standardized methodology, containing data from 89 drugs and 4867 datasets.
• Twenty-four EFs were extracted from electrical signals of four GI tissue types before and after drug treatment.
• ML models, including Naïve Bayes, SVM, and ensemble methods, were trained and validated to classify AEs.
• Extracted EFs were normalized and merged with the SIDER database for AE information.

Main Results:

• Nine ML models for AE classification were constructed with prediction accuracies ranging from 67% to 80%.
• Electrical features were categorized into 'excitatory' and 'inhibitory' types, correlating with specific AEs.
• Drugs were successfully clustered based on their EF profiles for the first time, revealing similarities among drugs acting on similar receptors.
• The study demonstrated the potential of GIPADD to predict drug targets.

Conclusions:

• GIPADD serves as a valuable big data resource for AI-driven drug discovery and AE prediction.
• Electrical features of GI pacemaker activity provide novel insights into drug safety and mechanisms of action.
• The developed ML approach offers a promising tool for identifying potential drug targets and improving drug safety assessments.