Predicting lncRNA and disease associations with graph autoencoder and noise robust gradient boosting

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Disease Mechanism Research

Background:

• Long non-coding RNAs (lncRNAs) play crucial roles in human diseases.
• Accurate identification of lncRNA-disease associations (LDAs) is vital for understanding disease mechanisms, discovering biomarkers, and improving diagnosis and treatment.
• Existing methods for LDA prediction require enhancement for improved accuracy and robustness.

Purpose of the Study:

• To develop and validate a novel computational framework, LDA-GARB, for predicting lncRNA-disease associations.
• To improve the accuracy and reliability of LDA prediction through advanced feature extraction and machine learning techniques.
• To provide a valuable tool for researchers investigating the roles of lncRNAs in human diseases.

Main Methods:

• LDA-GARB integrates nonnegative matrix factorization for linear feature extraction and a graph autoencoder for nonlinear feature extraction of lncRNAs and diseases.
• It computes lncRNA and disease similarities to enhance feature representation.
• A noise-robust gradient boosting model is employed to predict potential LDAs from combined features.

Main Results:

• LDA-GARB demonstrated superior performance in LDA prediction compared to several state-of-the-art methods across various cross-validation experiments.
• The framework showed robustness on imbalanced datasets and sensitivity analysis confirmed the effectiveness of its components.
• LDA-GARB successfully predicted potential lncRNA associations for colorectal cancer and breast cancer, identifying CCDC26 and HAR1A respectively.

Conclusions:

• LDA-GARB is an effective and robust computational tool for predicting lncRNA-disease associations.
• The framework contributes to advancing the understanding of lncRNA functions in disease pathogenesis.
• LDA-GARB offers a valuable resource for identifying novel disease biomarkers and therapeutic targets.