RPmirDIP: Reciprocal Perspective improves miRNA targeting prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• MicroRNAs (miRNAs) are key regulators of gene expression, interacting with messenger RNA (mRNA).
• Predicting miRNA-mRNA interactions is vital for understanding cellular functions.
• Existing prediction tools generate large-scale interaction datasets, such as mirDIP.

Purpose of the Study:

• To enhance the accuracy of miRNA-mRNA target prediction.
• To introduce a novel machine learning approach for improved interaction scoring.
• To facilitate the discovery of novel miRNA-gene interactions.

Main Methods:

• Application of the Reciprocal Perspective (RP) method, a semi-supervised machine learning technique.
• Development of RPmirDIP, adapting RP for miRNA-gene interaction prediction.
• Leveraging local thresholds from complementary views of miRNA-gene pairs.

Main Results:

• Significant improvement in miRNA target prediction accuracy ([Formula: see text]).
• RPmirDIP outperforms traditional global thresholding methods.
• Augmented prediction scores provide a more refined view of miRNA-gene interactions.

Conclusions:

• RPmirDIP offers a powerful new tool for miRNA target prediction.
• The method enhances the identification of biologically relevant miRNA-gene interactions.
• A comprehensive dataset of RPmirDIP-scored interactions is publicly available.