Computer-aided discovery of CDK16 inhibitors: a docking-augmented machine learning regression modelling approach

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Cyclin-dependent kinase 16 (CDK16) is crucial for cell cycle and implicated in aggressive cancers.
• Overexpression of CDK16 correlates with poor prognosis in various cancers.
• Lack of selective CDK16 inhibitors necessitates innovative drug discovery approaches.

Purpose of the Study:

• To develop a novel computational framework for identifying CDK16 inhibitors.
• To apply data augmentation using multiple docked poses for regression-based machine learning models.
• To discover novel and potent CDK16 inhibitors for potential anticancer therapies.

Main Methods:

• A computational framework integrating docked pose augmentation with regression-based machine learning (ML) was developed.
• Input features included docking scores, ligand-receptor contact fingerprints (LRCFs), and physicochemical descriptors.
• ML-QSAR models were optimized using Gradient Boosted Trees and the Genetic Function Algorithm for descriptor selection.
• Virtual screening was performed on NCI and OpnMe databases using a derived pharmacophore model.

Main Results:

• Gradient Boosted Trees identified as the optimal ML algorithm for predicting CDK16 inhibition.
• A validated ML-QSAR model guided the generation of a robust pharmacophore model.
• Two novel and potent CDK16 inhibitors, H_28 (IC50 = 3.5 µM) and H_33 (IC50 = 5.8 µM), were identified through in vitro assays.
• The study demonstrated the efficacy of docked pose augmentation for drug discovery.

Conclusions:

• The developed computational framework effectively integrates docked pose augmentation with regression modeling for accelerated drug discovery.
• The identified inhibitors represent promising leads for targeted anticancer agent development.
• This approach offers a robust strategy for discovering novel inhibitors against understudied kinases like CDK16.