mePTL: miRNA-encoded peptides identification using multi-source feature transformation and ensemble learning.

Area of Science:

• Bioinformatics
• Molecular Biology
• Genomics

Background:

• MicroRNAs (miRNAs) regulate gene expression post-transcriptionally.
• Small peptides, termed miPEPs, are translated from plant primary miRNA (pri-miRNA) small open reading frames (sORFs).
• While miPEPs are functionally diverse, computational methods for identifying plant miPEPs are limited.

Purpose of the Study:

• To introduce mePTL, a novel computational method for identifying plant miPEPs.
• To address the challenge of limited existing methods for plant miPEP identification.

Main Methods:

• Utilized multi-source feature transformation and ensemble learning.
• Extracted rich representation information from class-imbalanced miPEP and sORF data.
• Employed a feature representation learning framework and fused outputs from multiple machine learning models.

Main Results:

• mePTL demonstrated superior performance compared to existing methods on independent test sets.
• The method achieved high accuracy and strong generalization ability in miPEP identification.
• Experimental validation confirmed the effectiveness of the mePTL approach.

Conclusions:

• mePTL provides an effective and accurate computational approach for identifying plant miPEPs.
• The developed method can aid in the discovery of novel functional miPEPs.
• This work offers a valuable reference for future research in plant ncRNA and peptide discovery.