LRGCPND

LRGCPND predicts associations between non-coding RNAs and drug resistance by applying a linear residual graph convolution approach to model bipartite ncRNA–drug resistance relationships.

Key Features:

• Graph-Based Bipartite Construction: Constructs a bipartite graph using verified association data between non-coding RNAs (ncRNAs) and drug resistance.
• Linear Residual Graph Convolution: Aggregates features of neighboring nodes within each graph convolutional layer and transforms them via linear functions across layers to preserve and propagate biological signals.
• Layer Embedding Integration: Unites embedding representations from multiple layers to form comprehensive node representations for prediction.
• Neighborhood Feature Aggregation: Performs initial aggregation of features from neighboring nodes within the graph structure.
• Quantitative Performance: Achieved an average Area Under the Curve (AUC) of 0.8987 in comparative experiments against seven other state-of-the-art approaches.

Scientific Applications:

• ncRNA–Drug Resistance Association Prediction: Predicts candidate associations between non-coding RNAs and drug resistance to support discovery of resistance-associated ncRNAs.
• Mechanistic Investigation: Supports uncovering insights into mechanisms of drug action and resistance mediated by ncRNAs.
• Therapeutic Strategy Prioritization: Informs optimization of therapeutic strategies and clinical treatment planning by prioritizing ncRNA–drug resistance associations.

Methodology:

Constructs a bipartite graph from verified ncRNA–drug resistance association data; aggregates features from neighboring nodes within graph convolutional layers; applies linear transformations across layers as a linear residual graph convolution; and unites embeddings from multiple layers to generate final predictions.