GCRFLDA: scoring lncRNA-disease associations using graph convolution matrix completion with conditional random field.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long noncoding RNAs (lncRNAs) are crucial regulators in biological processes and disease development.
• Identifying lncRNA-disease associations aids in understanding disease mechanisms and developing treatments.
• Computational prediction of these associations complements experimental methods.

Purpose of the Study:

• To propose a novel computational method, GCRFLDA, for predicting lncRNA-disease associations.
• To leverage graph convolutional matrix completion for enhanced prediction accuracy.
• To validate the method's efficacy using benchmark datasets and case studies.

Main Methods:

• Constructed a graph from known lncRNA-disease associations.
• Employed a graph convolutional network with conditional random field and attention mechanisms for node embedding.
• Integrated Gaussian interaction profile kernel similarity and cosine similarity as node features.
• Utilized a decoder layer for scoring potential lncRNA-disease associations.

Main Results:

• GCRFLDA demonstrated superior performance compared to existing methods on four benchmark datasets.
• Case studies confirmed 70 out of 80 predicted lncRNA-disease associations through literature validation.
• The method effectively predicts novel associations, aiding biological discovery.

Conclusions:

• GCRFLDA is a powerful and accurate computational tool for predicting lncRNA-disease associations.
• The findings support the utility of computational approaches in advancing our understanding of lncRNA roles in disease.
• This method can accelerate the identification of potential therapeutic targets for various diseases.