Drug Repurposing Prediction for Immune-Mediated Cutaneous Diseases using a Word-Embedding-Based Machine Learning...

Area of Science:

• Bioinformatics and Computational Biology
• Immunology and Dermatology
• Pharmacology and Drug Discovery

Background:

• Immune-mediated diseases, particularly those affecting the skin, impact over 20% of the population.
• Drug repurposing offers a cost-effective strategy for identifying new therapeutic applications for existing medications.
• Autoimmune skin conditions like psoriasis, atopic dermatitis, and alopecia areata represent significant unmet medical needs.

Purpose of the Study:

• To develop and validate an efficient bioinformatics approach for drug repurposing in immune-mediated cutaneous diseases.
• To identify novel drug candidates for conditions such as psoriasis by modeling drug-disease relationships.
• To leverage machine learning and text summarization to predict effective drug repositioning strategies.

Main Methods:

• Utilized word embedding to process and summarize information from over 20 million scientific articles.
• Applied machine learning models to establish relationships between drugs and nine cutaneous and eight other immune-mediated diseases.
• Validated the approach through cross-validation, achieving a mean area under the receiver operating characteristic of 0.93.

Main Results:

• Successfully confirmed known effective drugs for psoriasis and identified novel potential drug candidates.
• Predicted drug targets were significantly enriched among genes differentially expressed in psoriatic skin lesions.
• The approach demonstrated high predictive performance across various immune-mediated diseases.

Conclusions:

• The developed bioinformatics algorithm provides a valuable tool for suggesting drug repurposing candidates for immune-mediated cutaneous diseases.
• While not determining clinical efficacy, the approach aids in prioritizing drugs for further investigation.
• This method facilitates the discovery of new treatments for challenging dermatological conditions.