GAM-MDR: probing miRNA-drug resistance using a graph autoencoder based on random path masking

Area of Science:

• Computational biology
• Genomics
• Pharmacology

Background:

• MicroRNAs (miRNAs) regulate gene expression and are key targets for disease therapies.
• Accurate prediction of miRNA-drug resistance (MDR) is essential for effective miRNA-based treatments.
• Deep learning models show promise for MDR prediction but can be hindered by data noise.

Purpose of the Study:

• To develop an advanced computational model for precise prediction of miRNA-drug resistance (MDR).
• To enhance the accuracy of miRNA-drug interaction predictions by mitigating data acquisition errors.

Main Methods:

• Introduction of the Graph Autoencoder with random path Masking for miRNA-drug resistance (GAM-MDR) model.
• Utilizing graph autoencoder (GAE) for efficient representation learning of miRNA and drug nodes.
• Implementing a random path masking strategy to reconstruct network paths and reduce noise impact.

Main Results:

• The GAM-MDR model demonstrated high reliability and effectiveness in predicting potential MDRs.
• The model successfully extracted robust representations of miRNA and drug nodes within the miRNA-drug network.
• Validation on public datasets confirmed the model's promising performance in MDR prediction.

Conclusions:

• The GAM-MDR model offers a novel approach to accurately predict miRNA-drug resistance.
• This method provides valuable insights for advancing miRNA therapeutic strategies and understanding miRNA regulatory mechanisms.
• The study highlights the potential of integrating graph autoencoders with random path masking for bioinformatics predictions.