Gtie-Rt: A comprehensive graph learning model for predicting drugs targeting metabolic pathways in human

Area of Science:

• Biochemistry
• Computational Biology
• Pharmacology

Background:

• Drugs modulate metabolic pathways for therapeutic outcomes, but pathway complexity complicates predicting overall metabolic effects.
• Understanding drug-metabolic pathway interactions is crucial for predicting drug efficacy and identifying potential drug-drug interactions.

Purpose of the Study:

• To develop and evaluate a hybrid machine learning model for accurately mapping drugs to their targeted metabolic pathways in humans.
• To enhance the understanding of drug-induced metabolic alterations and facilitate the prediction of drug-drug interactions.

Main Methods:

• Proposed a hybrid machine learning model, Graph Transformer Integrated Encoder (GTIE-RT).
• GTIE-RT integrates a Graph Convolutional Network (GCN) and a transformer encoder for graph embedding and attention.
• Utilized an Extremely Randomized Trees Classifier for predicting target metabolic pathways.

Main Results:

• The GTIE-RT model achieved excellent performance on a drug dataset.
• Key performance metrics included accuracy (>95%), recall (>92%), precision (>93%), and F1-score (>92%).
• GTIE-RT demonstrated superior and more reliable results compared to other machine learning methods and model variants.

Conclusions:

• The GTIE-RT model offers a robust and accurate approach for mapping drugs to human metabolic pathways.
• This computational tool can significantly improve the prediction of drug metabolic effects and aid in identifying potential drug-drug interactions.
• The model's high performance suggests its utility in drug discovery and personalized medicine.