Protein language model-based prediction for plant miRNA encoded peptides

Area of Science:

• Plant molecular biology
• Bioinformatics
• Peptide science

Background:

• Plant miRNA encoded peptides (miPEPs) regulate essential plant traits, but their identification is hindered by limited training data.
• Current prediction methods like miPEPPred-FRL rely on manually engineered features, limiting their scope and accuracy.

Purpose of the Study:

• To develop an advanced computational model for accurate plant miPEP identification.
• To leverage deep learning for improved feature representation in peptide sequence analysis.

Main Methods:

• Integration of ESM2 peptide embedding with machine learning algorithms.
• Development and validation of the pLM4PEP prediction model.

Main Results:

• pLM4PEP achieves high accuracy in identifying plant miPEPs.
• The model demonstrates robust performance across diverse datasets, including other bioactive peptides.
• Successful application of deep learning for enhanced peptide feature extraction.

Conclusions:

• pLM4PEP offers a powerful and accurate tool for plant miPEP discovery.
• The approach advances the field of bioactive peptide prediction using deep learning.
• The model's versatility extends to identifying other types of bioactive peptides.