Machine Learning and Deep Learning Approaches for Antimicrobial Activity Prediction Using Molecular Descriptors

Abstract

The rapid increase in antimicrobial resistance has created an urgent demand for efficient methods to discover novel antimicrobial compounds. Traditional laboratory screening methods are expensive, time consuming, and often infeasible for evaluating large chemical libraries. Computational approaches using machine learning and deep learning have emerged as promising alternatives for accelerating antimicrobial discovery. This work explores machine learning and deep learning approaches for antimicrobial activity prediction, leveraging molecular descriptor data for both classification and MIC prediction tasks. Several predictive pipelines were implemented, including XGBoost-based machine learning models, deep neural networks, and autoencoder-based feature compression techniques. Both classification and regression tasks were performed to predict minimum inhibitory concentration (MIC) values. Experimental results show that XGBoost achieves the highest classification accuracy for antimicrobial activity prediction, while the autoencoder-enhanced deep neural network provides more balanced precision, recall, and F1-score values. For MIC prediction, XGBoost regression provides the most accurate predictions. The comparative analysis highlights the complementary strengths of deep learning and tree-based ensemble models for computational antimicrobial screening.