Deep-Learning vs Physics-Based Docking Tools for Future Coronavirus Pandemics.

Area of Science:

• Computational chemistry
• Drug discovery
• Artificial intelligence in medicine

Background:

• The COVID-19 pandemic highlighted the need for rapid antiviral drug discovery.
• Evaluating AI's role in accelerating the development of drugs against future coronavirus outbreaks is crucial.

Purpose of the Study:

• To assess the effectiveness of three AI-driven molecular docking approaches in accelerating drug discovery.
• To compare physics-based, deep learning-assisted, and data-driven AI methods for antiviral drug discovery.

Main Methods:

• Utilized SARS-CoV-2 and MERS-CoV datasets for evaluating molecular docking.
• Compared AutoDock Vina (physics-based), GNINA (deep learning-assisted), and Boltz-2 (data-driven) docking methods.
• Assessed AI model performance based on docking precision and prediction of binding potency.

Main Results:

• The data-driven Boltz-2 model achieved over 80% accuracy, significantly enhancing docking precision.
• Accurate prediction of binding potency remains a challenge, necessitating further methodological development.
• Traditional methods (Vina, GNINA) offer speed for large-scale screening and easier interpretability.

Conclusions:

• AI has significantly transformed the drug discovery landscape, improving precision and speed.
• Integrating AI tools into accessible platforms can decentralize drug discovery.
• Further research is needed to enhance AI's ability to predict binding potency accurately.