Virtual Screening of Small Molecules Targeting BCL2 with Machine Learning, Molecular Docking, and MD Simulation

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• The BCL-2 protein plays a crucial role in apoptosis and is a target for cancer therapy.
• Identifying small molecule inhibitors for BCL-2 is essential for developing novel anti-cancer drugs.

Purpose of the Study:

• To screen a large library of small molecules for potential BCL-2 inhibitors.
• To utilize machine learning algorithms and molecular simulations for efficient drug discovery.

Main Methods:

• Employed deep neural networks (DNN) and random forest (RF) algorithms for initial screening.
• Utilized molecular docking with Vina to assess binding affinity.
• Performed molecular dynamics (MD) simulations to validate compound stability and binding.

Main Results:

• The RF model identified 61% of screened molecules as potentially active.
• Identified three top-scoring compounds: CHEMBL3940231, CHEMBL3938023, and CHEMBL3947358, with docking scores of -11, -10.9, and -10.8 kcal/mol.
• MD simulations confirmed the stability and binding affinity of these compounds to the BCL2 protein.

Conclusions:

• The study successfully identified promising small molecule inhibitors of BCL-2.
• The combination of AI and molecular simulations provides an effective strategy for drug discovery.
• These identified compounds warrant further investigation for therapeutic potential against BCL-2 related diseases.