Deep Probabilistic Matrix Factorization on Graphs: Application to Drug Repositioning in Antimicrobial Resistance

Area of Science:

• Computational biology
• Pharmacology
• Infectious diseases

Background:

• Antimicrobial resistance (AMR) is a growing global health crisis.
• Misuse and overuse of antibiotics lead to treatment failures.
• Effective treatments for resistant infections are urgently needed.

Purpose of the Study:

• To develop a novel algorithm for general-purpose drug repositioning.
• To address the challenge of ineffective first-line antibiotics due to AMR.
• To identify potential antibiotic treatments for resistant bacterial infections.

Main Methods:

• Developed a new algorithm for drug repositioning using a matrix completion framework on graphs.
• Employed a probabilistic approach combining deep matrix factorization and graph learning.
• Curated a new dataset of antibiotic-bacteria associations.

Main Results:

• The proposed method demonstrated superior performance compared to benchmarks.
• The algorithm achieved precise drug repurposing for general applications.
• Successfully applied to three specific antimicrobial resistance (AMR) case studies.

Conclusions:

• The developed algorithm offers a promising approach for drug repositioning in the context of AMR.
• This method can aid in identifying effective treatments for infections caused by resistant bacteria.
• The findings contribute to combating the global health challenge of antimicrobial resistance.