Deep Learning-Based Potential Ligand Prediction Framework for COVID-19 with Drug-Target Interaction Model.

Area of Science:

• Computational drug discovery
• Virology
• Machine learning in medicine

Background:

• The COVID-19 pandemic necessitates rapid development of effective antiviral drugs.
• Identifying potent drug candidates that bind to the SARS-CoV-2 S-glycoprotein is crucial for therapeutic intervention.
• Existing drug discovery methods require significant time and resources.

Purpose of the Study:

• To computationally screen and identify ligands with high binding affinity to the SARS-CoV-2 S-glycoprotein.
• To develop and validate a machine learning model for predicting drug-target interactions.
• To propose a list of potential drug candidates for COVID-19 treatment.

Main Methods:

• Implementation of a 1D convolutional neural network architecture for predicting drug-target interaction (DTI) values.
• Training the network on the KIBA (Kinase Inhibitor Bioactivity) dataset to learn binding affinity patterns.
• Predicting KIBA scores for a diverse set of ligands against the 2019-nCoV S-glycoprotein.

Main Results:

• The 1D convolutional network successfully predicted KIBA scores, indicating binding affinity.
• A list of 33 top-performing ligands demonstrating high binding affinity to the S-glycoprotein was generated.
• These identified ligands represent promising candidates for novel coronavirus drug development.

Conclusions:

• The study successfully identified potential drug candidates for COVID-19 through computational screening.
• The developed 1D convolutional network model is effective for predicting drug-target interactions and accelerating drug discovery.
• The proposed ligands warrant further experimental validation for their therapeutic potential against SARS-CoV-2.