A Support Vector Machine based method to distinguish proteobacterial proteins from eukaryotic plant proteins

Area of Science:

• Agricultural Science
• Bioinformatics
• Machine Learning

Background:

• Proteobacteria bacteria cause significant plant diseases, reducing agricultural yield and quality.
• Early detection of plant infections is crucial for mitigating crop losses.
• Next-generation sequencing and mass spectrometry enable macromolecule-based plant screening.

Purpose of the Study:

• To develop a systematic method for determining the organismal origin of peptide fragments.
• To create a machine learning model capable of distinguishing between proteobacterial and plant proteins.

Main Methods:

• Utilized a Support Vector Machine (SVM) algorithm.
• Developed models based on amino acid composition (AAC) and dipeptide composition (DC).
• Implemented a hybrid approach combining AAC and DC for enhanced accuracy.

Main Results:

• Proteobacterial proteins exhibit distinct amino acid compositions compared to plant proteins.
• The AAC-based SVM model achieved 92.44% accuracy and 0.85 MCC.
• The DC-based SVM model reached 94.67% accuracy and 0.89 MCC.
• The hybrid AAC and DC SVM model demonstrated a maximum accuracy of 94.86% and 0.90 MCC.
• Models were validated on unseen datasets.

Conclusions:

• The hybrid SVM model effectively distinguishes proteobacterial from plant protein sequences.
• This approach aids in early identification of plant diseases caused by Proteobacteria.