Autoencoder-based drug-virus association prediction with reliable negative sample selection: A case study with...

Area of Science:

• Virology
• Computational Biology
• Drug Discovery

Background:

• Emerging and re-emerging viruses pose significant global health and economic challenges.
• Identifying novel virus-drug associations is crucial for combating viral threats and developing effective treatments.
• Traditional experimental methods for drug discovery are costly and time-consuming.

Purpose of the Study:

• To propose a computational method, KR-AEVDA, for predicting and prioritizing novel virus-drug associations.
• To leverage machine learning for efficient identification of potential antiviral drugs.
• To reduce the cost and risk associated with experimental drug discovery.

Main Methods:

• KR-AEVDA utilizes k-nearest neighbor (KNN) for reliable negative sample selection.
• Autoencoder-based feature extraction is employed to analyze complex drug-virus relationships.
• An ensemble classifier infers novel virus-drug associations based on extracted features and similarity data.

Main Results:

• KR-AEVDA demonstrated superior performance compared to existing state-of-the-art methods across three datasets.
• Molecular docking validated the top predicted drug associations with SARS-CoV-2's main protease.
• Case studies, particularly for SARS-CoV-2, highlighted the method's effectiveness in identifying potential therapeutic strategies.

Conclusions:

• KR-AEVDA offers a robust and efficient computational approach for discovering virus-drug associations.
• The method aids in prioritizing drug candidates for experimental validation, accelerating the drug discovery pipeline.
• KR-AEVDA shows promise in addressing the challenges posed by emerging viral infections.