MGATRx: Discovering Drug Repositioning Candidates Using Multi-View Graph Attention

Area of Science:

• Computational biology
• Drug discovery and development
• Bioinformatics

Background:

• In-silico drug repositioning is a cost-effective strategy for identifying new therapeutic indications for existing drugs.
• Challenges in drug repositioning include the heterogeneity and sparseness of biological data, such as disease and drug annotations.
• Existing similarity-based methods often struggle with the complexity of these biological networks.

Purpose of the Study:

• To propose a novel multi-view graph attention network for indication discovery (MGATRx).
• To leverage integrated drug-centric and disease-centric annotations to improve drug repositioning accuracy.
• To address the limitations of current methods in handling complex and sparse biological data.

Main Methods:

• Developed a multi-view graph attention network (MGATRx) integrating drug and disease data.
• Employed graph attention mechanisms to learn node representations within a heterogeneous information network.
• Utilized both drug-centric and disease-centric annotations as distinct data views.

Main Results:

• MGATRx demonstrated superior performance compared to four state-of-the-art computational drug repositioning methods.
• Several novel drug indications predicted by MGATRx are currently under investigation or supported by existing literature.
• The model's predictions highlight its potential translational utility in drug discovery.

Conclusions:

• MGATRx offers an effective computational approach for in-silico drug repositioning and indication discovery.
• The multi-view graph attention network framework successfully integrates diverse biological data for enhanced prediction accuracy.
• The validated predictions underscore the practical applicability of MGATRx in accelerating the identification of new drug uses.