Predicting genes associated with RNA methylation pathways using machine learning.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• RNA methylation is crucial for regulating RNA function and is a growing area of interest in biological research and drug discovery.
• Understanding the full spectrum of genes involved in RNA methylation pathways is essential for advancing these fields.

Purpose of the Study:

• To predict novel genes associated with RNA methylation pathways in humans using integrated omics data.
• To identify molecular sub-networks that connect predicted genes to known RNA methylation processes.

Main Methods:

• Collected and integrated transcriptomic, proteomic, structural, and physical interaction data from the Harmonizome database.
• Applied supervised machine learning algorithms, including five types of classifiers, trained and evaluated using cross-validation.
• Utilized protein-protein interaction data to construct molecular sub-networks.

Main Results:

• Achieved high prediction accuracy, with cross-validation reaching 88% and test set accuracy averaging 91%.
• Identified six molecular sub-networks linking predicted genes to known RNA methylation genes.
• These networks are implicated in mRNA methylation, tRNA and rRNA processing, and protein/chromatin modifications.

Conclusions:

• Machine learning applied to large omics datasets is a powerful approach for predicting gene function, specifically in the context of RNA methylation pathways.
• The identified sub-networks provide new insights into the complex regulatory roles of RNA methylation beyond direct RNA modification.