Targeting in silico GPCR conformations with ultra-large library screening for hit discovery

Area of Science:

• Computational Biology
• Structural Biology
• Pharmacology

Background:

• Deep machine learning (ML) provides numerous protein conformations, potentially aiding structure-based drug discovery (SBDD) where experimental data is scarce.
• The utility of ML-predicted conformations for screening vast chemical libraries against protein targets remains uncertain.

Purpose of the Study:

• To evaluate the suitability of ML-predicted protein conformations for ultra-large library screening (ULLS).
• To discuss the benefits and limitations of using these conformations in drug discovery, particularly for dynamic targets.

Main Methods:

• Review and analysis of current deep machine learning models for protein structure prediction.
• Examination of the challenges posed by ultra-large libraries (ULLs) in drug screening.
• Consideration of protein conformational dynamics in drug target selection.

Main Results:

• ML models generate accessible annotated protein conformations, offering a potential alternative to experimental structures.
• The accuracy of predicted conformations for effective small molecule binding requires further investigation.
• Targeting diverse conformational states, such as those in G-protein-coupled receptors (GPCRs), may enhance drug discovery efforts.

Conclusions:

• ML-predicted protein structures show promise for SBDD and ULLS.
• Further validation is needed to confirm the reliability of predicted conformations for pharmacological screening.
• Exploiting protein conformational heterogeneity, particularly for GPCRs, presents a promising avenue for developing novel therapeutics.