An Effective Plant Small Secretory Peptide Recognition Model Based on Feature Correction Strategy

Area of Science:

• Plant molecular biology
• Bioinformatics
• Computational biology

Background:

• Plant small secretory peptides (SSPs) regulate crucial biological processes.
• Accurate SSP identification is vital for functional studies.
• Experimental validation is accurate but time-consuming and costly.
• Existing machine learning methods face limitations due to feature extraction instability.

Purpose of the Study:

• To develop an advanced, accurate, and efficient computational tool for plant SSP prediction.
• To address the limitations of traditional methods and current machine learning approaches in SSP recognition.

Main Methods:

• Proposed a novel feature-correction-based model named SE-SSP.
• Utilized transformer encoders to capture implicit sequence features.
• Introduced a 2-D SENET module for adaptive feature correction and robust representation.
• Employed stacked linear modules for deep feature extraction.
• Implemented a contrastive learning strategy to handle sparse sample data.

Main Results:

• The SE-SSP model demonstrated excellent performance in predicting plant SSPs.
• Experimental results on public datasets validated the model's effectiveness.
• The feature correction and transformer encoder components significantly improved prediction accuracy.

Conclusions:

• The SE-SSP model offers a convenient and effective tool for plant SSP prediction.
• This approach advances computational methods in plant science.
• The model's public availability facilitates further research and application.