LRSSL: predict and interpret drug-disease associations based on data integration using sparse subspace learning.

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Identifying novel drug indications is vital for drug development.
• Integrating multiple data types aids in discovering new drug uses.
• Efficiently linking drugs to diseases and understanding mechanisms remains challenging.

Purpose of the Study:

• To present a novel method for predicting indications of both new and approved drugs.
• To efficiently identify mechanisms behind drug-disease associations by integrating diverse data.

Main Methods:

• Laplacian regularized sparse subspace learning (LRSSL) was developed.
• The method integrates drug chemical, target domain, and target annotation information.
• L1-norm constraint was used for extracting important drug features.

Main Results:

• The proposed LRSSL method outperforms existing approaches in predicting drug-disease associations.
• Predicted therapeutic effects were validated through database records and literature.
• Extracted drug features aid in interpreting prediction results.

Conclusions:

• The LRSSL method offers a robust approach for drug indication prediction.
• The method enhances understanding of drug-disease mechanisms.
• Extracted features provide valuable insights for drug repurposing and development.