Machine Learning Study of Metabolic NetworksvsChEMBL Data of Antibacterial Compounds

Abstract

Antibacterial drugs (AD) change the metabolic status of bacteria, contributing to bacterial death. However, antibiotic resistance and the emergence of multidrug-resistant bacteria increase interest in understanding metabolic network (MN) mutations and the interaction of AD vs MN. In this study, we employed the IFPTML = Information Fusion (IF) + Perturbation Theory (PT) + Machine Learning (ML) algorithm on a huge dataset from the ChEMBL database, which contains >155,000 AD assays vs >40 MNs of multiple bacteria species. We built a linear discriminant analysis (LDA) and 17 ML models centered on the linear index and based on atoms to predict antibacterial compounds. The IFPTML-LDA model presented the following results for the training subset: specificity (Sp) = 76% out of 70,000 cases, sensitivity (Sn) = 70%, and Accuracy (Acc) = 73%. The same model also presented the following results for the validation subsets: Sp = 76%, Sn = 70%, and Acc = 73.1%. Among the IFPTML nonlinear models, the k nearest neighbors (KNN) showed the best results with Sn = 99.2%, Sp = 95.5%, Acc = 97.4%, and Area Under Receiver Operating Characteristic (AUROC) = 0.998 in training sets. In the validation series, the Random Forest had the best results: Sn = 93.96% and Sp = 87.02% (AUROC = 0.945). The IFPTML linear and nonlinear models regarding the ADs vs MNs have good statistical parameters, and they could contribute toward finding new metabolic mutations in antibiotic resistance and reducing time/costs in antibacterial drug research.

Multi-Endpoint Acute Toxicity Assessment of Organic Compounds Using Large-Scale Machine Learning Modeling

“…The study revealed that using the multicondition descriptor along with the K-nearest neighbors model showed the best results with sensitivity = 99.2%, specificity = 95.5%, and accuracy = 97.4%, elaborating on the capability of the QSAR model to classify the antibacterial drugs correctly. 38 The majority of prior studies that utilized MCDs focused on classification tasks. This technique is implemented in the current work as well, where we will employ MCDs to perform a regression task for predicting acute toxicity.…”

“…When the end points are mechanistically correlated, a single general model can be used to incorporate information from several end points to predict the toxicity. , This will be more reliable than using separate models for each end point, as it can consider the potential interactions between end points that are not apparent when we investigate them in isolation. Multicondition descriptors (MCDs) are a helpful tool for combining data from various assays to create a coherent source of data for creating single-task models. , Multicondition descriptors, by their very nature, allow us to create a more robust and accurate model from multiple data sources . In 2021, Dieguez-Santana and González-Dı́az developed a QSAR model using multicondition descriptors that could hasten the design of Dual Antibacterial Drug Nanoparticles (DADNP) systems formed by antibacterial drugs and nanoparticles.…”

“…The minimal bactericide concentration (MBC), the specific protein, the assay organism in the experiment, the metabolic network microorganism species, and the target type were all different conditions under which the data set was collected. The study revealed that using the multicondition descriptor along with the K-nearest neighbors model showed the best results with sensitivity = 99.2%, specificity = 95.5%, and accuracy = 97.4%, elaborating on the capability of the QSAR model to classify the antibacterial drugs correctly . The majority of prior studies that utilized MCDs focused on classification tasks.…”

“…Multicondition descriptors (MCDs) are a helpful tool for combining data from various assays to create a coherent source of data for creating single-task models. 37,38 Multicondition descriptors, by their very nature, allow us to create a more robust and accurate model from multiple data sources. 39 In 2021, Dieguez-Santana and Gonzaĺez-Dı ́az developed a QSAR model using multi-condition descriptors that could hasten the design of Dual Antibacterial Drug Nanoparticles (DADNP) systems formed by antibacterial drugs and nanoparticles.…”

Artificial intelligence and machine learning in Leishmania drug discovery

“…QSAR modelling tools have been used for the identification of potential drug candidates [14]. They have evolved into AI-based QSAR approaches, such as linear discriminant analysis (LDA), support vector machines (SVM), neural networks (NN), random forests (RF), and decision trees, which can be applied to accelerate QSAR analysis [15][16][17][18]. AI/ML can be used successfully in drug discovery, including drug design, polypharmacology, chemical synthesis, drug repurposing, and drug screening (reproduced from [19,20]).…”