Predicting Inhibition of CDK2 with SAnDReS: The Application of Machine Learning to Navigate the Scoring Function...

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Cyclin-dependent kinase 2 (CDK2) is crucial for cell cycle progression and a key target for anticancer therapies.
• Extensive structural data on CDK2 over 30 years enables the development of computational models for studying its inhibition.
• Machine learning (ML) models are increasingly utilized to predict drug-target interactions.

Purpose of the Study:

• To review the application of computational models, specifically ML, for estimating CDK2 inhibition.
• To evaluate the predictive performance of ML models built using the SAnDReS program for CDK2 inhibitors.
• To compare SAnDReS-generated ML models against classical and deep-learning scoring functions.

Main Methods:

• Literature search on PubMed for ML models predicting CDK2 inhibition.
• Utilizing BindingDB data for CDK2 to generate updated ML models.
• Applying the SAnDReS program to model CDK2-inhibitor interactions and analyze predictive performance using DOME analysis.

Main Results:

• ML models built with SAnDReS demonstrate superior predictive performance for CDK2 inhibition (pKi) compared to classical and deep-learning scoring functions.
• DOME analysis confirms the suitability of ML models for predicting protein-ligand interactions.
• The study validates the use of SAnDReS for developing high-performance predictive models for CDK2 inhibitors.

Conclusions:

• Rich structural and functional data of CDK2 supports the development of robust ML models for predicting its inhibition.
• SAnDReS offers a powerful computational approach to build superior ML models for predicting pKi, outperforming existing scoring functions.
• This work highlights the potential of ML and SAnDReS in advancing the discovery of CDK2-targeting anticancer drugs.