Stable solution to l _2,1-based robust inductive matrix completion and its application in linking long noncoding RNAs...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Numerous long intergenic non-coding RNAs (lincRNAs) are implicated in human diseases, but many associations remain unconfirmed.
• Experimental validation of lincRNA-disease links is costly and time-consuming.
• High-throughput sequencing and GWAS have generated vast lincRNA data, enabling computational approaches for relationship discovery.

Purpose of the Study:

• To develop an in silico tool for inferring lincRNA-disease associations.
• To address limitations of existing methods, such as inability to handle noisy data and sparsity.
• To simultaneously utilize side information for both lincRNAs and diseases within a unified framework.

Main Methods:

• Proposed Stable Robust Inductive Matrix Completion (SRIMC), a novel technique based on Inductive Matrix Completion (IMC).
• Employed l2,1 norm-based regularization for objective function optimization.
• Implemented a unique 2-step stable solution approach to enhance robustness.

Main Results:

• SRIMC demonstrated superior performance over state-of-the-art methods in precision@k and recall@k for lincRNA-disease prioritization.
• The method effectively predicted associations for novel lincRNAs.
• SRIMC accurately ranked newly identified diseases for known lincRNAs.

Conclusions:

• SRIMC robustly handles datasets with noise and outliers.
• The developed method is effective for discovering associations involving novel lincRNAs and disease phenotypes.
• SRIMC offers a powerful computational approach for advancing lincRNA-disease association research.