Deep variational graph autoencoders for novel host-directed therapy options against COVID-19

Area of Science:

• Computational Biology
• Drug Discovery
• Artificial Intelligence in Medicine

Background:

• The COVID-19 pandemic necessitates urgent development of therapeutic options.
• Repurposing existing drugs offers a rapid implementation pathway due to prior approval.
• Advanced artificial intelligence (AI) techniques for drug screening remain largely underexplored.

Purpose of the Study:

• To identify novel therapeutic targets and drug repurposing candidates for COVID-19 using AI.
• To explore the potential of deep learning techniques for predicting drug-host-pathogen interactions.
• To establish new host-directed therapy (HDT) options against SARS-CoV-2.

Main Methods:

• Constructed a comprehensive molecular interaction network integrating drug-protein and protein-protein interactions with SARS-CoV-2 protein data.
• Employed variational graph autoencoders (VGAEs), a deep learning method, to learn network structure and predict missing links.
• Focused on novel drug-human protein links crucial for SARS-CoV-2 replication.

Main Results:

• Identified previously unknown links between drugs and human proteins vital for SARS-CoV-2 replication.
• Predicted novel host-directed therapy (HDT) options with high plausibility confirmed by simulations.
• Highlighted potential drug targets essential for viral survival that can be addressed by existing medications.

Conclusions:

• Advanced AI, specifically VGAEs, can effectively predict drug-host-pathogen interactions for COVID-19.
• The study presents a network-based approach to discover novel HDT strategies.
• Predicted interactions offer promising avenues for repurposing existing drugs to combat SARS-CoV-2.