ABDpred: Prediction of active antimicrobial compounds using supervised machine learning techniques.

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• The rise of multidrug-resistant pathogens necessitates novel antibiotic discovery.
• Conventional methods face low success rates, highlighting the need for advanced approaches.
• In silico methods, particularly machine learning, offer a promising alternative for identifying new antimicrobial agents.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting novel antimicrobial compounds.
• To address the bottleneck in antibiotic discovery caused by the increasing threat of resistant pathogens.
• To create an accessible tool for researchers to aid in the identification of potential new antibiotics.

Main Methods:

• Eight machine learning algorithms were employed: extreme gradient boosting, random forest, gradient boosting classifier, deep neural network, support vector machine, multilayer perceptron, decision tree, and logistic regression.
• Models were trained on a dataset of 312 known antibiotic drugs and 936 non-antibiotic compounds.
• A five-fold cross-validation approach was used to train and evaluate the ML models.

Main Results:

• The top four ML classifiers (extreme gradient boosting, random forest, gradient boosting classifier, and deep neural network) achieved accuracies of 80% and above.
• These high accuracies were consistent across both testing and blind datasets.
• An ensemble model was created by aggregating the top four classifiers using a soft-voting technique.

Conclusions:

• The developed ensemble ML model demonstrates significant potential for predicting novel antimicrobial compounds.
• The ensemble model, integrated into the ABDpred online server, provides a freely accessible resource for antibiotic discovery.
• This computational approach offers a promising strategy to accelerate the identification of new antibiotics against resistant pathogens.