This study proposes a method to predict lncRNA-drug associations (LDAs) using graph convolutional networks (GCN) and graph attention networks (GAT). The method integrates lncRNA and drug similarities after principal component analysis (PCA) denoising as node attributes in the lncRNA-drug pair subgraphs. The model leverages the inherent graph structures of LDA networks and similarity networks to selectively focus on important local information and integrate global information. The proposed method achieves high performance on five datasets, with average AUCs exceeding 0.92. Case studies on two drugs (Berberine and Panobinostat) and two lncRNAs (NEAT1 and MEG3) demonstrate the model's effectiveness in predicting potential LDAs. Functional enrichment analysis further validates the predicted LDAs by examining the relationship between the biological functions of drugs and the enrichment pathways of lncRNA target genes. Overall, the study provides a deep learning-based framework for predicting novel LDAs, accelerating the development of lncRNA-targeted therapies.This study proposes a method to predict lncRNA-drug associations (LDAs) using graph convolutional networks (GCN) and graph attention networks (GAT). The method integrates lncRNA and drug similarities after principal component analysis (PCA) denoising as node attributes in the lncRNA-drug pair subgraphs. The model leverages the inherent graph structures of LDA networks and similarity networks to selectively focus on important local information and integrate global information. The proposed method achieves high performance on five datasets, with average AUCs exceeding 0.92. Case studies on two drugs (Berberine and Panobinostat) and two lncRNAs (NEAT1 and MEG3) demonstrate the model's effectiveness in predicting potential LDAs. Functional enrichment analysis further validates the predicted LDAs by examining the relationship between the biological functions of drugs and the enrichment pathways of lncRNA target genes. Overall, the study provides a deep learning-based framework for predicting novel LDAs, accelerating the development of lncRNA-targeted therapies.