Machine learning applications in RNA modification sites prediction.

Area of Science:

• Computational Biology
• Molecular Biology
• Bioinformatics

Background:

• Post-transcriptional chemical changes in Ribonucleic acid (RNA) modifications are vital regulators of RNA function.
• Advances in deep sequencing and large-scale profiling have generated extensive transcriptomic datasets.
• These datasets facilitate the application of machine learning in epitranscriptomics for identifying RNA modifications.

Purpose of the Study:

• To comprehensively review machine learning-based approaches for predicting 11 distinct types of RNA modifications.
• To analyze the entire workflow of machine learning methods for RNA modification site prediction.
• To compare existing methods based on datasets, species, approaches, and accuracy.

Main Methods:

• Exploration of benchmark datasets used in RNA modification prediction.
• Analysis of feature extraction techniques relevant to RNA sequences.
• Review of various classification algorithms applied to epitranscriptomic data.

Main Results:

• Comparison of machine learning methods across different RNA modification types, highlighting variations in datasets, target species, and predictive accuracy.
• Identification of common strategies and challenges in applying machine learning to epitranscriptomics.

Conclusions:

• Machine learning offers powerful tools for predicting RNA modifications, driven by increasing data availability.
• The review elucidates the strengths and weaknesses of current computational approaches.
• Future perspectives focus on advancing prediction accuracy and expanding the scope of RNA modification identification.