Gene Ontology-based function prediction of long non-coding RNAs using bi-random walk.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Advancements in sequencing technology have identified numerous long non-coding RNAs (lncRNAs).
• While some lncRNAs are known to regulate development via dosage compensation and epigenetic mechanisms, most lack functional characterization.
• Determining lncRNA functions and regulatory networks is a current research priority.

Purpose of the Study:

• To develop a scalable network-based model for predicting long non-coding RNA functions.
• To create a comprehensive model integrating lncRNA similarity, lncRNA-protein associations, and protein-protein interactions.
• To annotate lncRNAs with Gene Ontology (GO) terms based on their network neighbors.

Main Methods:

• Construction of a global network integrating lncRNA similarity, lncRNA-protein association, and protein-protein interaction (PPI) networks.
• Application of a bi-random walk algorithm to identify similarities between lncRNAs and proteins.
• Utilizing neighboring protein information for lncRNA functional annotation with GO terms.

Main Results:

• BiRWLGO was compared against state-of-the-art models using a manually curated lncRNA benchmark dataset.
• The model demonstrated superior performance, achieving higher maximum F-measure (Fmax) and coverage.
• Integration of protein interaction data significantly enhanced BiRWLGO's predictive accuracy.

Conclusions:

• BiRWLGO provides an efficient method for predicting lncRNA functions at scale.
• The inclusion of protein interaction data substantially improves the predictive power of the model.
• This approach facilitates the functional characterization of unannotated lncRNAs.