BPAGS: a web application for bacteriocin prediction via feature evaluation using alternating decision tree, genetic...

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Antibiotic resistance necessitates novel therapeutic strategies.
• Bacteriocins offer a promising alternative to conventional antibiotics.
• Accurate prediction of novel bacteriocins is crucial for drug development.

Purpose of the Study:

• To develop a precise and efficient computational model for predicting novel bacteriocins.
• To create a user-friendly web application for bacteriocin prediction.
• To compare machine learning approaches with existing prediction tools.

Main Methods:

• Feature extraction from physicochemical, structural, and sequence-profile attributes.
• Feature selection using Alternating Decision Tree (ADTree), Genetic Algorithm (GA), and Linear Support Vector Classifier (linear SVC).
• Model construction using Random Forest (RF), Support Vector Machine (SVM), Decision Tree (DT), Logistic Regression (LR), k-Nearest Neighbors (KNN), and Gaussian Naïve Bayes (GNB).

Main Results:

• The Support Vector Machine (SVM) model with ADTree-reduced features achieved the highest accuracy (99.11%) and AUC (0.9984).
• The developed web application, BPAGS, integrates predictive models based on ADTree, GA, and linear SVC.
• Comparative analysis showed superior performance of the developed models over sequence alignment and deep learning approaches.

Conclusions:

• Machine learning, particularly SVM with ADTree-selected features, provides a highly accurate method for bacteriocin prediction.
• The BPAGS web application offers a valuable tool for identifying novel bacteriocins.
• This approach aids in the development of new drugs to combat antibiotic resistance.