A Computational Method for the Identification of Endolysins and Autolysins

Area of Science:

• Biochemistry
• Computational Biology
• Microbiology

Background:

• Cell lytic enzymes offer a promising alternative to antibiotics for treating bacterial infections due to their ability to destroy bacterial structures.
• Unlike antibiotics, cell lytic enzymes do not contribute to the growing problem of drug resistance in pathogenic bacteria.
• Distinguishing between endolysins and autolysins is essential for understanding their roles in cell wall lysis and developing targeted therapies.

Purpose of the Study:

• To develop an efficient computational method for predicting the type of cell lytic enzyme (endolysin or autolysin).
• To overcome the time-consuming nature of experimental methods for cell lytic enzyme identification.
• To provide a valuable tool for research into bacterial infection treatments.

Main Methods:

• A dataset of 27 endolysins and 41 autolysins was compiled.
• Protein sequences were represented using tripeptide composition for feature extraction.
• Feature selection was performed based on confidence degree.
• A Support Vector Machine (SVM) classifier was trained for prediction.

Main Results:

• The proposed computational method achieved an overall accuracy of 97.06% with 44 selected features.
• This represents a significant improvement of nearly 4.5% compared to existing methods.
• The method demonstrated stable performance across a range of feature numbers (40-70).
• The tripeptide optimal feature set yielded 94.12% accuracy, outperforming Chou's amphiphilic PseAAC method by approximately 18%.

Conclusions:

• An efficient computational method for identifying endolysins and autolysins was successfully developed.
• The Support Vector Machine classifier, utilizing selected features, demonstrated superior performance.
• The proposed method offers a highly accurate and reliable approach for cell lytic enzyme type prediction.