Optimized network based natural language processing approach to reveal disease comorbidities in COVID-19

Area of Science:

• Computational biology
• Network medicine
• Virology

Background:

• COVID-19 pandemic highlighted the need to understand comorbidities for better healthcare.
• Traditional methods for identifying comorbidities lack mechanistic insights.
• Comorbidity prediction is crucial for prioritizing care and developing therapies.

Purpose of the Study:

• To discover unknown COVID-19 comorbidities using a deep learning approach.
• To gain mechanistic insights into disease connections through miRNA regulatory interactions.
• To integrate multi-disease data for enhanced comorbidity detection.

Main Methods:

• Utilized a modified deep learning algorithm (mpDisNet) based on word-embedding techniques.
• Incorporated miRNA expression profiles from SARS-CoV-2 infected cells and their target transcription factors.
• Applied network medicine principles to connect diseases via miRNA-mediated regulatory interactions.

Main Results:

• The algorithm successfully predicted most known COVID-19 comorbidities.
• Identified several potentially novel comorbidities associated with COVID-19.
• Demonstrated the capability to uncover diseases linked through miRNA regulatory pathways.

Conclusions:

• Deep learning offers a powerful approach to uncover mechanistically-informed comorbidities.
• The identified novel comorbidities warrant further investigation for awareness and prevention.
• This method advances network medicine and aids preparedness for future outbreaks.