iPADD: A Computational Tool for Predicting Potential Antidiabetic Drugs Using Machine Learning Algorithms.

Area of Science:

• Computational Chemistry
• Pharmacology
• Bioinformatics

Background:

• Diabetes mellitus is a chronic metabolic disease with increasing prevalence, necessitating novel therapeutic agents.
• Current drug discovery for diabetes is time-consuming and resource-intensive.
• Artificial intelligence (AI) offers a powerful approach to accelerate the identification of potential antidiabetic drugs.

Purpose of the Study:

• To develop and validate a predictive model, iPADD, for discovering novel antidiabetic drugs.
• To leverage machine learning and molecular descriptors for efficient drug screening.
• To assess the generalization ability of the developed model through independent testing and case analysis.

Main Methods:

• Utilized four types of molecular fingerprints and their combinations to encode drug molecules.
• Employed minimum-redundancy-maximum-relevance (mRMR) and incremental feature selection for optimal feature screening.
• Trained and evaluated eight machine learning algorithms using 5-fold cross-validation.
• Validated the predictor's performance on an independent test set and through molecular docking studies.

Main Results:

• The best machine learning model achieved an accuracy of 0.983 and an auROC of 0.989 on the independent test set.
• iPADD correctly predicted the antidiabetic potential of 64 out of 65 natural products analyzed.
• Molecular docking confirmed stable binding of quercetin and resveratrol to the diabetes target NR1I2, aligning with model predictions.

Conclusions:

• The iPADD predictor, based on machine learning, demonstrates high accuracy and strong generalization ability for identifying potential antidiabetic drugs.
• AI-driven approaches significantly enhance the efficiency of antidiabetic drug discovery.
• The developed model provides a valuable tool for researchers seeking novel therapeutic compounds for diabetes management.