A Quantum-Inspired Method for Three-Dimensional Ligand-Based Virtual Screening.

Area of Science:

• Computational Chemistry
• Quantum Computing
• Drug Discovery

Background:

• Molecular similarity assessment is crucial for virtual screening (VS) in early-stage drug discovery.
• Current methods often rely on molecular fingerprints, which may not capture all structural nuances.
• There is a need for advanced similarity methods that can incorporate diverse molecular features.

Purpose of the Study:

• To introduce a novel quantum-inspired graph-based molecular similarity (GMS) method.
• To evaluate the GMS method's performance in ligand-based VS, incorporating pharmacophore features and 3D atomic coordinates.
• To compare the GMS method against conventional fingerprint-based approaches.

Main Methods:

• Formulated the GMS method as a quadratic unconstrained binary optimization (QUBO) problem.
• Utilized quantum annealing for solving the QUBO problem.
• Incorporated pharmacophore features and 3D atomic coordinates into the GMS model.
• Evaluated performance on datasets from the DUD_LIB_VS_1.0 library.

Main Results:

• The GMS method using 3D atomic coordinates demonstrated higher early enrichment values.
• The GMS approach outperformed traditional fingerprint methods across most tested datasets.
• Quantum annealing offers a novel computational approach for molecular similarity calculations.

Conclusions:

• The quantum-inspired GMS method presents a promising alternative for ligand-based virtual screening.
• Incorporating 3D atomic coordinates enhances the effectiveness of molecular similarity assessments.
• This approach has the potential for significant advancements in drug discovery pipelines.