Circumventing the Synthesizability Problem in Generative Molecular Design.

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Generative structure-based drug design (SBDD) models accelerate the discovery of novel drug candidates.
• A key limitation of current SBDD models is the generation of compounds with poor synthesizability, hindering practical drug design.
• Existing virtual ligand screening (VLS) methods face challenges with increasingly large chemical spaces.

Purpose of the Study:

• To address the synthesizability issue in generative SBDD.
• To introduce a novel model-guided virtual screening (MGVS) pipeline.
• To enhance the efficiency and applicability of SBDD in drug discovery.

Main Methods:

• Developed an MGVS pipeline integrating SBDD models with chemical similarity search.
• Paired SBDD models with efficient search methods to identify synthesizable analogues in ultralarge compound databases.
• Validated the approach across diverse protein targets and three SBDD models.

Main Results:

• Reliably identified synthesizable analogues with comparable or superior docking scores and binding poses.
• MGVS demonstrated a consistent 25x improvement in docking-based screening efficiency compared to standard VLS.
• The approach proved effective across a wide range of protein targets.

Conclusions:

• MGVS effectively overcomes the synthesizability limitations of generative SBDD models.
• MGVS significantly enhances screening efficiency, offering a practical solution for drug discovery.
• This approach is crucial for navigating vast chemical spaces and advancing future drug discovery efforts.