PEA-m6A: an ensemble learning framework for accurately predicting N6-methyladenosine modifications in plants.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• N 6-methyladenosine (m6A) is the most abundant mRNA modification in eukaryotes, regulating gene expression and RNA metabolism.
• Accurate m6A prediction is crucial for understanding its biological roles, but existing models are often species-specific and limited in scope.

Purpose of the Study:

• To develop a unified, flexible, and accurate framework for predicting m6A-modified regions in plant genomes.
• To address the limitations of existing species-centric m6A prediction models.

Main Methods:

• Developed PEA-m6A, a modularized and parameterized framework utilizing ensemble learning.
• Integrated statistic-based and deep learning-driven features for m6A prediction.
• Employed transfer learning to enhance prediction accuracy with small sample sizes.

Main Results:

• PEA-m6A demonstrated superior performance compared to the state-of-the-art predictor WeakRM, with AUC improvements of 6.7% to 23.3% across 12 plant species.
• The framework showed strong generalization capabilities for both within- and cross-species m6A prediction.
• PEA-m6A effectively leverages knowledge from pre-trained models, improving accuracy in low-sample scenarios.

Conclusions:

• PEA-m6A provides a highly accurate, flexible, and transferable tool for m6A prediction in plants.
• The framework's generalization ability makes it suitable for diverse plant genomic applications.
• Packaged with Galaxy and Docker for user-friendliness, PEA-m6A is publicly available to advance m6A research.