Predicting lncRNA-Protein Interaction With Weighted Graph-Regularized Matrix Factorization.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Long non-coding RNAs (lncRNAs) play crucial roles in biological activities, often through interactions with proteins.
• Experimental validation of lncRNA-protein interactions (LPIs) is resource-intensive and time-consuming.
• Understanding LPIs is essential for elucidating lncRNA functions.

Purpose of the Study:

• To develop an accurate computational method for predicting unobserved lncRNA-protein interactions (LPIs).
• To leverage lncRNA and protein similarity data along with known interactions for prediction.
• To provide a more efficient alternative to experimental validation methods.

Main Methods:

• Developed a weighted graph-regularized matrix factorization (LPI-WGRMF) method.
• Utilized lncRNA similarity matrices, protein similarity matrices, and known LPIs as input.
• Compared LPI-WGRMF against five established LPI prediction methods (LPBNI, LPI-IBNRA, LPIHN, RWR, CF).

Main Results:

• LPI-WGRMF demonstrated high-accuracy performance in predicting LPIs.
• Achieved an Area Under the Curve (AUC) score of 0.9012 and an Area Under the Precision-Recall Curve (AUPR) of 0.7324.
• Identified potential novel interactions, including SFPQ with Q9NUL5, SNHG3 with Q9NUL5, and PRPF31 with Q9UKV8, requiring experimental verification.

Conclusions:

• The LPI-WGRMF method is effective for predicting lncRNA-protein interactions.
• This computational approach can significantly accelerate the discovery of functional LPIs.
• The predicted interactions warrant further experimental validation to confirm biological significance.