Garbage in, garbage out: how reliable training data improved a virtual screening approach against SARS-CoV-2 MPro.

Area of Science:

• Computational chemistry and drug discovery
• Virology and infectious diseases
• Biochemistry and enzymology

Background:

• SARS-CoV-2 replication inhibition is a key research priority.
• Computational methods accelerate drug discovery but require reliable data and validation.
• The SARS-CoV-2 main protease (MPro) is a critical target for antiviral development.

Purpose of the Study:

• To identify novel chemical compounds inhibiting SARS-CoV-2 MPro using computational and experimental approaches.
• To refine computational models for drug discovery through iterative learning cycles.
• To validate the efficacy of identified compounds against SARS-CoV-2 replication.

Main Methods:

• Employed a drug discovery strategy combining in silico virtual screening with experimental validation.
• Utilized ligand-based and structure-based computational approaches for virtual screening of a large chemolibrary.
• Conducted iterative refinement of computational models based on experimental screening results and published data.

Main Results:

• Initial screening identified three MPro inhibitors, including a glycoside, a benzothiazole, and a flavonol.
• Second-generation models yielded 43 new hit candidates; eight compounds inhibited MPro (IC₅₀ 0.12–20 μM).
• Five of the eight MPro inhibitors also demonstrated antiviral activity against SARS-CoV-2 in cell culture (EC₅₀ 7–45 μM).

Conclusions:

• Demonstrated a successful iterative drug discovery process integrating computational predictions and experimental validation.
• Confirmed the 'garbage in, garbage out' principle in machine learning for drug discovery.
• Identified promising lead compounds for further development as SARS-CoV-2 antivirals.