lncRNA-disease association prediction method based on the nearest neighbor matrix completion model

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long noncoding RNAs (lncRNAs) are implicated in various diseases, but their large-scale detection is challenging and costly.
• Efficient methods are needed to screen and identify disease-associated lncRNAs, improving biological experiment efficiency.

Purpose of the Study:

• To develop a novel prediction model for lncRNA-disease associations using the nearest neighbor concept.
• To introduce a new similarity algorithm that fuses potential associations for enhanced prediction accuracy.

Main Methods:

• Implementation of a nearest neighbor-based prediction model for lncRNA-disease association.
• Development and application of a novel similarity algorithm integrating potential associations.
• Validation using leave-one-out cross-validation and a new independent dataset.

Main Results:

• The proposed model demonstrated superior performance compared to Cosine, Pearson, and Jaccard similarity algorithms.
• Leave-one-out cross-validation achieved an Area Under the Curve (AUC) of 0.96, indicating excellent predictive power.
• Validation on a new dataset yielded an AUC of 0.92, confirming the model's reliability and generalizability.

Conclusions:

• The developed nearest neighbor model offers an efficient and reliable method for predicting lncRNA-disease associations.
• This approach can significantly aid in the screening of disease-related lncRNAs, reducing experimental costs and improving efficiency.
• The novel similarity algorithm is a key component contributing to the model's high predictive accuracy.