Computational prediction of type III secreted proteins from gram-negative bacteria

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Type III secretion system (T3SS) is crucial for bacterial virulence in plant and animal pathogens.
• T3SS is also vital in rhizobia for symbiosis, but secretion mechanisms remain unclear.
• Identifying type III secreted effectors (T3SEs) is challenging due to lack of defined signals.

Purpose of the Study:

• Develop a bioinformatics approach to predict novel T3SEs.
• Identify T3SEs in plant pathogens and rhizobia.

Main Methods:

• Utilized a machine learning model trained on N-terminal amino acid sequences.
• Features included amino acid composition, secondary structure, and solvent accessibility.
• Combined with a promoter screen for T3SS-specific promoters.

Main Results:

• Achieved over 90% precision in predicting T3SEs in *Pseudomonas syringae*.
• Identified 57 candidate T3SEs in four rhizobial strains.
• Confirmed 17 of these candidates as actual effectors.

Conclusions:

• Machine learning based on N-terminal sequences and promoter screening is effective for predicting T3SEs.
• This computational approach aids in identifying bacterial virulence factors.