Graph Neural Network with Self-Supervised Learning for Noncoding RNA-Drug Resistance Association Prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Noncoding RNA (ncRNA) plays a crucial role in drug resistance, making its association with drug response a significant area for drug development.
• Experimental methods for identifying ncRNA-drug resistance links are often slow and limited in scale.
• There is an urgent need for efficient computational approaches to predict these associations.

Purpose of the Study:

• To develop a computational framework, GSLRDA, for predicting associations between ncRNA and drug resistance.
• To leverage graph convolutional networks and self-supervised learning for improved prediction accuracy.
• To provide a scalable and efficient tool for identifying potential ncRNA-drug resistance relationships.

Main Methods:

• Constructing a bipartite graph representing known ncRNA-drug resistance associations.
• Employing light graph convolutional network (lightGCN) to learn vector representations of ncRNA and drug nodes.
• Utilizing data augmentation and contrastive self-supervised learning to enhance node representations.
• Predicting associations using the inner product of learned vector representations.

Main Results:

• GSLRDA achieved a high Area Under the Curve (AUC) of 0.9101, surpassing eight other state-of-the-art models.
• The framework demonstrated effectiveness in predicting ncRNA-drug resistance associations.
• Case studies further validated the predictive power of GSLRDA for specific drugs.

Conclusions:

• GSLRDA represents a novel application of self-supervised learning in bioinformatics for association prediction.
• The developed framework offers a significant advancement in predicting ncRNA-drug resistance associations.
• GSLRDA provides a valuable tool for accelerating drug discovery and development by identifying key ncRNA players in drug resistance.