RFMirTarget: predicting human microRNA target genes with a random forest classifier

Area of Science:

• Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial regulators of eukaryotic gene expression, impacting numerous cellular pathways.
• Accurate identification of miRNA targets is a significant bioinformatics challenge, driving the development of computational methods.
• Machine learning approaches have shown superior specificity and sensitivity in miRNA target prediction.

Purpose of the Study:

• To introduce and evaluate RFMirTarget, a novel microRNA-target prediction method utilizing a random forest classifier.
• To explore the efficacy of random forest algorithms in the specific context of miRNA target prediction.
• To assess the performance of RFMirTarget against existing state-of-the-art methods.

Main Methods:

• RFMirTarget analyzes miRNA-target alignments, extracting features related to structure, thermodynamics, sequence alignment, seed regions, and binding position.
• A random forest classifier is employed for the classification of miRNA-target pairs based on these extracted features.
• Feature selection benefits were investigated, even with the classifier's inherent feature importance analysis.

Main Results:

• RFMirTarget demonstrated statistically significant outperformance compared to several established classifiers.
• The method's performance was robust against class imbalance and feature correlation issues.
• Comparative analysis on independent datasets (TarBase, starBase) revealed promising performance, notably higher sensitivity than other methods.
• Feature selection confirmed its utility and highlighted consistency with biologically relevant properties.

Conclusions:

• RFMirTarget represents a significant advancement in computational miRNA target prediction, offering enhanced sensitivity.
• The random forest approach is well-suited for miRNA target prediction, addressing key challenges in the field.
• The study underscores the importance of feature selection for optimizing predictive models in bioinformatics.