Lorenc et. al., Hybrid Machine Learning Approach for Predicting MIC of Imidazole Chlorides Using CART, PCA, and a Single Neural Network: Dataset

This repository contains datasets and documentation used in the research focused on predicting antimicrobial activity of bis-imidazolium compounds using a hybrid machine learning approach. The files support preprocessing, feature selection, and modeling steps involving molecular descriptors. Contents 1. Lorenc_imidazole_CART_PCA_NN_Saureus_Calbicans_dragon_molecular_descriptor_list.pdfA reference file listing all molecular descriptors generated by the DRAGON software, including descriptor names, descriptions, and the corresponding molecular blocks. This file provides interpretability and traceability for descriptor meanings used in the analysis. 2. Lorenc_imidazole_CART_PCA_NN_Saureus_Calbicans_full_initial_dataset.csvThe full dataset used in the research, consisting of over 5,000 molecular descriptors computed using the DRAGON software. Each row represents a bis-imidazolium compound, and columns correspond to various calculated molecular features. 3. Lorenc_imidazole_CART_PCA_NN_Saureus_Calbicans_PCs.csvA dataset containing the selected principal components (PCs) derived from dimensionality reduction of the original molecular descriptors. These PCs were used as input features in the final predictive model.