Improved QSAR models for PARP-1 inhibition using data balancing, interpretable machine learning, and matched...

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• Poly (ADP-ribose) polymerase-1 (PARP-1) is a key target in breast cancer therapy.
• Existing PARP-1 inhibitors like Olaparib can cause severe hematological and cardiotoxicity.
• Current in silico models for PARP-1 activity prediction suffer from low specificity and interpretability.

Purpose of the Study:

• To develop a machine learning (ML)-based quantitative structure-activity relationship (QSAR) model for predicting PARP-1 activity.
• To enhance the interpretability of predictive models for PARP-1 inhibitors.
• To guide the design of novel PARP-1 inhibitors with improved efficacy and safety profiles.

Main Methods:

• A comprehensive ML-based QSAR approach was employed.
• Classification models utilized the Synthetic Minority Oversampling Technique (SMOTE) and K-nearest neighbor algorithm.
• Mechanistic interpretation was achieved using Shapley Additive Explanations (SHAP) and Matched Molecular Pair Analysis (MMPA).

Main Results:

• Robust classification models achieved 0.86 accuracy, 0.88 sensitivity, and 0.80 specificity.
• Regression models showed strong performance on specific compound classes like phthalazinones.
• SHAP analysis identified key topological features, and MMPA extracted chemical transformation rules.

Conclusions:

• The developed ML-QSAR approach provides accurate and interpretable predictions of PARP-1 activity.
• The findings offer mechanistic insights into PARP-1 inhibition.
• The study provides valuable guidance for medicinal chemists in designing next-generation PARP-1 inhibitors for breast cancer treatment.