Predicting Cell Wall Lytic Enzymes Using Combined Features.

Area of Science:

• Biochemistry
• Computational Biology
• Microbiology

Background:

• Antibiotic resistance is a growing global health threat.
• Cell wall lytic enzymes offer a potential alternative to conventional antibiotics.
• Accurate prediction of cell wall lytic enzymes is crucial for developing new antimicrobial strategies.

Purpose of the Study:

• To propose an improved computational method for predicting cell wall lytic enzymes.
• To enhance the accuracy and reliability of cell wall lytic enzyme identification.
• To provide a valuable tool for protein function prediction in the context of antimicrobial resistance.

Main Methods:

• Utilized amino acid composition (AAC), dipeptide composition (DC), position-specific score matrix auto-covariance (PSSM-AC), and auto-covariance average chemical shift (acACS) as features.
• Employed Support Vector Machine (SVM) for classification.
• Applied Synthetic Minority Over-sampling Technique (SMOTE) to address imbalanced datasets and F-score for feature selection.

Main Results:

• Achieved high prediction accuracy with S = 99.35%, S = 99.02%, MCC = 0.98, and Acc = 99.19% using the optimized feature combination (AAC+DC+acACS+PSSM-AC).
• The developed predictive model demonstrated superior performance compared to existing methods.
• The jackknife test confirmed the robustness and effectiveness of the proposed method.

Conclusions:

• The improved prediction method for cell wall lytic enzymes shows significant potential.
• This approach can aid in the discovery of novel antimicrobial agents.
• The method contributes to advancing protein function prediction and combating antimicrobial resistance.