Prediction of RNA Methylation Status From Gene Expression Data Using Classification and Regression Methods.

Area of Science:

• Molecular Biology
• Epigenetics
• Bioinformatics

Background:

• RNA N⁶-methyladenosine (m⁶A) is a crucial epigenetic modification regulating RNA stability, processing, and translation.
• Aberrant m⁶A homeostasis is linked to stem cell dysfunction, infertility, and cancer.
• Experimental m⁶A detection is laborious, and matched epitranscriptome data is scarce.

Purpose of the Study:

• To investigate the feasibility of predicting RNA m⁶A methylation status from gene expression data using computational approaches.
• To develop and evaluate predictive models for m⁶A status based on available gene expression datasets.

Main Methods:

• Utilized mouse RNA methylation data from 73 experimental conditions.
• Employed machine learning algorithms including Elastic Net-regularized Logistic Regression (ENLR), Support Vector Machine (SVM), and Random Forests (RF) for classification.
• Applied Elastic Net for regression analysis.

Main Results:

• SVM and RF classifiers achieved a mean Area Under the Curve (AUC) of 0.84, indicating good predictive performance.
• Gene Site Enrichment Analysis identified significant associations with phosphoprotein, SH3 domain, and endoplasmic reticulum, all relevant to the m⁶A pathway.
• Regression models yielded a mean Pearson correlation of 0.68 and a mean Spearman correlation of 0.64.

Conclusions:

• Gene expression data can be effectively used to construct accurate predictors for RNA m⁶A methylation status.
• This computational approach facilitates m⁶A research, particularly in scenarios lacking direct RNA methylation profiles.