Decoding potential lncRNA and disease associations through graph representation learning and gradient boosting with...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long noncoding RNAs (lncRNAs) play crucial roles in complex diseases.
• Experimental validation of lncRNA-disease associations (LDAs) is resource-intensive.
• Accurate prediction of LDAs is vital for disease understanding and therapeutic development.

Purpose of the Study:

• To develop an efficient and accurate computational model for inferring lncRNA-disease associations (LDAs).
• To leverage advanced machine learning techniques for enhanced LDA prediction.
• To provide a valuable tool for identifying potential lncRNA biomarkers for complex diseases.

Main Methods:

• The LDA-GMCB model integrates graph embedding learning, multi-head self-attention (MSA) with convolutional neural networks (CNN), low-rank singular value decomposition (SVD), and histogram-based gradient boosting (HGBoost).
• Nonlinear features are captured using graph embedding and MSA-CNN, while linear features are extracted via low-rank SVD.
• HGBoost is employed for the final inference of lncRNA-disease relationships.

Main Results:

• LDA-GMCB demonstrated superior performance compared to four baseline models and four popular classifiers across 5-fold cross-validation and cold-start scenarios.
• The model achieved significant improvements on the lncRNADisease and MNDR databases.
• Ablation studies confirmed the effectiveness of individual components within LDA-GMCB.

Conclusions:

• LDA-GMCB offers a robust and efficient computational approach for predicting lncRNA-disease associations.
• The model successfully identified potential lncRNAs (DGCR5, HIF1A) associated with Alzheimer's and Parkinson's diseases.
• LDA-GMCB provides a valuable resource for future research into lncRNA functions in complex diseases.