Predicting A-to-I RNA editing by feature selection and random forest.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genetics

Background:

• RNA editing is a crucial post-transcriptional process in eukaryotes, enhancing RNA and protein diversity.
• Computational prediction of RNA editing remains a significant challenge in molecular biology.

Purpose of the Study:

• To develop a novel computational method for predicting RNA editing.
• To identify optimal features for accurate RNA editing prediction.

Main Methods:

• A random forest model was employed for RNA editing prediction.
• Feature selection was performed using Maximum Relevance Minimum Redundancy (mRMR) and Incremental Feature Selection (IFS) algorithms.
• An optimal set of 18 features was selected from an initial set of 77 features.

Main Results:

• The developed predictor achieved high accuracy (0.866 training, 0.876 testing) and Matthews correlation coefficient (0.742 training, 0.576 testing).
• A reduced feature set of 18 features provided superior predictive performance compared to using all 77 features.
• Analysis of the selected features offers insights into RNA editing mechanisms.

Conclusions:

• A robust random forest-based method for RNA editing prediction has been established.
• A concise set of features is sufficient for accurate RNA editing prediction.
• The identified features provide a foundation for further experimental investigation into RNA editing factors.