DPPPredIV

DPPPredIV integrates binary classification and continuous regression Quantitative Structure Activity Relationships (QSAR) models to predict human dipeptidyl peptidase 4 (DPP4) inhibitory activity for early-stage drug discovery screening.

Key Features:

• Binary Classification: Soft voting ensemble of multiple machine learning classification models predicts DPP4 inhibitor activity.
• Continuous Regression: Regression models estimate IC50 values for predicted inhibitors.
• Descriptor Selection: Genetic algorithm selects molecular descriptors from compound data.
• Applicability Domain Analysis: All models undergo applicability domain analysis to ensure prediction reliability.
• SMILES Input: Predictions are generated from SMILES string inputs.
• External Validation: Models undergo external 10-fold cross-validation.

Scientific Applications:

• Type 2 Diabetes Mellitus Drug Discovery: Identifies novel DPP4 inhibitors to accelerate early-stage screening and reduce costs in T2DM therapeutic development.
• Virtual Screening: Experimental testing of 29 MolPort compounds at 1.5 µM confirmed 14 predicted actives exhibited significant inhibition, supporting performance in compound prioritization.

Methodology:

A curated dataset of 4,676 ChEMBL compounds was subjected to genetic algorithm descriptor selection and multiple machine learning algorithms; classification models were combined via soft voting ensemble, while regression models estimated IC50 values.