Machine learning and molecular dynamics simulations predict potential TGR5 agonists for type 2 diabetes treatment

Area of Science:

• Computational chemistry and cheminformatics
• Pharmacology and drug discovery
• Metabolic diseases research

Background:

• Type 2 diabetes (T2D) treatment is challenging due to complex metabolic pathways.
• Takeda G protein-coupled receptor 5 (TGR5) is a promising target for T2D therapy, regulating glucose homeostasis and energy expenditure.
• TGR5 agonists can improve glycemic control, making them attractive therapeutic candidates.

Purpose of the Study:

• To identify novel small molecules as potential TGR5 agonists for T2D treatment.
• To explore the utility of machine learning (ML), molecular docking (MD), and molecular dynamics simulations (MDS) in drug discovery.
• To predict and validate compounds with high affinity and stability for TGR5.

Main Methods:

• Collected bioactivity data for known TGR5 agonists from the ChEMBL database.
• Developed a Random Forest ML model using molecular descriptors and screened the COCONUT database.
• Performed molecular docking and 100 ns molecular dynamics simulations on top-scoring compounds.

Main Results:

• The ML model predicted potential TGR5 agonists from the COCONUT database.
• Molecular docking indicated that lead compounds possess stronger affinity for TGR5 than the co-crystallized ligand.
• Molecular dynamics simulations confirmed the binding stability of the lead compounds within the TGR5 active site.

Conclusions:

• The integrated approach of ML, MD, and MDS is effective for discovering novel TGR5 agonists.
• Identified lead compounds demonstrate potential for optimization in T2D therapeutic strategies.
• This study highlights a powerful computational strategy for identifying new drug candidates for metabolic diseases.