Mycobacterium tuberculosis Cell Wall Permeability Model Generation Using Chemoinformatics and Machine Learning Approaches

Free Energy Barriers for Passive Drug Transport through the Mycobacterium tuberculosis Outer Membrane: A Molecular Dynamics Study

The emergence of multi-drug-resistant tuberculosis strains poses a significant challenge to modern medicine. The development of new antituberculosis drugs is hindered by the low permeability of many active compounds through the extremely strong bacterial cell wall of mycobacteria. In order to estimate the ability of potential antimycobacterial agents to diffuse through the outer mycolate membrane, the free energy profiles, the corresponding activation barriers, and possible permeability modes of passive transport for a series of known antibiotics, modern antituberculosis drugs, and prospective active drug-like molecules were determined using molecular dynamics simulations with the all-atom force field and potential of mean-force calculations. The membranes of different chemical and conformational compositions, density, thickness, and ionization states were examined. The typical activation barriers for the low-mass molecules penetrating through the most realistic membrane model were 6-13 kcal/mol for isoniazid, pyrazinamide, and etambutol, and 19 and 25 kcal/mol for bedaquilin and rifampicin. The barriers for the ionized molecules are usually in the range of 37-63 kcal/mol. The linear regression models were derived from the obtained data, allowing one to estimate the permeability barriers from simple physicochemical parameters of the diffusing molecules, notably lipophilicity and molecular polarizability.

Maintenance of cell wall remodeling and vesicle production are connected in Mycobacterium tuberculosis

Pathogenic and nonpathogenic mycobacteria secrete extracellular vesicles (EVs) under various conditions. EVs produced by Mycobacterium tuberculosis ( Mtb ) have raised significant interest for their potential in cell communication, nutrient acquisition, and immune evasion. However, the relevance of vesicle secretion during tuberculosis infection remains unknown due to the limited understanding of mycobacterial vesicle biogenesis. We have previously shown that a transposon mutant in the LCP-related gene virR ( virR mut ) manifested a strong attenuated phenotype during experimental macrophage and murine infections, concomitant to enhanced vesicle release. In this study, we aimed to understand the role of VirR in the vesicle production process in Mtb . We employ genetic, transcriptional, proteomics, ultrastructural and biochemical methods to investigate the underlying processes explaining the enhanced vesiculogenesis phenomenon observed in the virR mutant. Our results establish that VirR is critical to sustain proper cell permeability via regulation of cell envelope remodeling possibly through the interaction with similar cell envelope proteins, which control the link between peptidoglycan and arabinogalactan. These findings advance our understanding of mycobacterial extracellular vesicle biogenesis and suggest that these set of proteins could be attractive targets for therapeutic intervention.

Machine learning based dynamic consensus model for predicting blood-brain barrier permeability

The blood-brain barrier (BBB) is an important defence mechanism that restricts disease-causing pathogens and toxins to enter the brain from the bloodstream. In recent years, many in silico methods were proposed for predicting BBB permeability, however, the reliability of these models is questionable due to the smaller and class-imbalance dataset which subsequently leads to a very high false positive rate. In this study, machine learning and deep learning-based predictive models were built using XGboost, Random Forest, Extra-tree classifiers and deep neural network. A dataset of 8153 compounds comprising both the BBB permeable and BBB non-permeable was curated and subjected to calculations of molecular descriptors and fingerprints for generating the features for machine learning and deep learning models. Three balancing techniques were then applied to the dataset to address the class-imbalance issue. A comprehensive comparison among the models showed that the deep neural network model generated on the balanced MACCS fingerprint dataset outperformed with an accuracy of 97.8% and a ROC-AUC score of 0.98 among all the models. Additionally, a dynamic consensus model was prepared with the machine learning models and validated with a benchmark dataset for predicting BBB permeability with higher confidence scores.

Assessing machine learning approaches for predicting failures of investigational drug candidates during clinical trials

One of the major challenges in drug development is having acceptable levels of efficacy and safety throughout all the phases of clinical trials followed by the successful launch in the market. While there are many factors such as molecular properties, toxicity parameters, mechanism of action at the target site, etc. that regulates the therapeutic action of a compound, a holistic approach directed towards data-driven studies will invariably strengthen the predictive toxicological sciences. Our quest for the current study is to find out various reasons as to why an investigational candidate would fail in the clinical trials after multiple iterations of refinement and optimization. We have compiled a dataset that comprises of approved and withdrawn drugs as well as toxic compounds and essentially have used time-split based approach to generate the training and validation set. Five highly robust and scalable machine learning binary classifiers were used to develop the predictive models that were trained with features like molecular descriptors and fingerprints and then validated rigorously to achieve acceptable performance in terms of a set of performance metrics. The mean AUC scores for all the five classifiers with the hold-out test set were obtained in the range of 0.66-0.71. The models were further used to predict the probability score for the clinical candidate dataset. The top compounds predicted to be toxic were analyzed to estimate different dimensions of toxicity. Apparently, through this study, we propose that with the appropriate use of feature extraction and machine learning methods, one can estimate the likelihood of success or failure of investigational drugs candidates thereby opening an avenue for future trends in computational toxicological studies. The models developed in the study can be accessed at https://github.com/gnsastry/predicting_clinical_trials.git.

Machine Learning Prediction of Mycobacterial Cell Wall Permeability of Drugs and Drug-like Compounds

The cell wall of Mycobacterium tuberculosis and related organisms has a very complex and unusual organization that makes it much less permeable to nutrients and antibiotics, leading to the low activity of many potential antimycobacterial drugs against whole-cell mycobacteria compared to their isolated molecular biotargets. The ability to predict and optimize the cell wall permeability could greatly enhance the development of novel antitubercular agents. Using an extensive structure-permeability dataset for organic compounds derived from published experimental big data (5371 compounds including 2671 penetrating and 2700 non-penetrating compounds), we have created a predictive classification model based on fragmental descriptors and an artificial neural network of a novel architecture that provides better accuracy (cross-validated balanced accuracy 0.768, sensitivity 0.768, specificity 0.769, area under ROC curve 0.911) and applicability domain compared with the previously published results.

Deciphering the importance of MD descriptors in designing Vitamin D Receptor agonists and antagonists using machine learning

The Vitamin D Receptor (VDR) ligand-binding domain undergoes conformation change upon the binding of VDR agonists/antagonists. Helix 12 ((H)12) is one of the important helices at VDR ligand binding and its conformational changes are controlled by the binding of agonists and antagonists molecules. Various molecular modeling studies are available to explain the agonistic and antagonistic activity of vitamin D analogs. In this work, for the first time, we attempted to generate a machine learning model with fingerprints, 2D, 3D and MD descriptors that are specific to Vitamin D analogs and VDR. Initially, 2D and 3D descriptors and fingerprints of 1003 vitamin D analogs were calculated using CDK and RDKit. The machine learning model was generated using descriptors and fingerprints. Further, 80 Vitamin D analogs (40 VDR agonists + 40 VDR antagonists) were docked in the VDR active site. 50ns MD simulation was performed for each protein-ligand complex. Different MD descriptors such as Solvent Accessible Surface Area (SASA), radius of gyration, PC1 and PC2 were calculated and considered along with CDK and RDKit descriptors as features for machine learning calculations. A few other descriptors that are related to VDR conformational changes such as conformation of the (H)12, the angle at kink were considered for machine learning model generation. It was observed that the descriptors calculated from VDR conformational changes i) were able to distinguish between agonists and antagonists ii) provide key and comprehensive information about the unique binding characteristics of agonists and antagonists iii) provide a strong basis for the machine learning model generation. Overall, this study attempts the utilization of descriptors that are specific to a protein conformation will be helpful for the generation of an efficient machine learning model.

Artificial intelligence in virtual screening: models versus experiments

A typical drug discovery project involves identifying active compounds with significant binding potential for selected disease-specific targets. Experimental high-throughput screening (HTS) is a traditional approach to drug discovery, but is expensive and time-consuming when dealing with huge chemical libraries with billions of compounds. The search space can be narrowed down with the use of reliable computational screening approaches. In this review, we focus on various machine-learning (ML) and deep-learning (DL)-based scoring functions developed for solving classification and ranking problems in drug discovery. We highlight studies in which ML and DL models were successfully deployed to identify lead compounds for which the experimental validations are available from bioassay studies.

Artificial intelligence: machine learning for chemical sciences

Research in molecular sciences witnessed the rise and fall of Artificial Intelligence (AI)/ Machine Learning (ML) methods, especially artificial neural networks, few decades ago. However, we see a major resurgence in the use of modern ML methods in scientific research during the last few years. These methods have had phenomenal success in the areas of computer vision, speech recognition, natural language processing (NLP), etc. This has inspired chemists and biologists to apply these algorithms to problems in natural sciences. Availability of high performance Graphics Processing Unit (GPU) accelerators, large datasets, new algorithms, and libraries has enabled this surge. ML algorithms have successfully been applied to various domains in molecular sciences by providing much faster and sometimes more accurate solutions compared to traditional methods like Quantum Mechanical (QM) calculations, Density Functional Theory (DFT) or Molecular Mechanics (MM) based methods, etc. Some of the areas where the potential of ML methods are shown to be effective are in drug design, prediction of high-level quantum mechanical energies, molecular design, molecular dynamics materials, and retrosynthesis of organic compounds, etc. This article intends to conceptually introduce various modern ML methods and their relevance and applications in computational natural sciences. Synopsis Recent surge in the application of machine learning (ML) methods in fundamental sciences has led to a perspective that these methods may become important tools in chemical science. This perspective provides an overview of the modern ML methods and their successful applications in chemistry during the last few years.

Drug repositioning for anti-tuberculosis drugs: an in silico polypharmacology approach

Development of potential antitubercular molecules is a challenging task due to the rapidly emerging drug-resistant strains of Mycobacterium tuberculosis (M.tb). Structure-based approaches hold greater benefit in identifying compounds/drugs with desired polypharmacological profiles. These methods can be employed based on the knowledge of protein binding sites to identify the complementary ligands. In this study, polypharmacology guided computational drug repurposing approach was applied to identify potential antitubercular drugs. 20 important druggable protein targets in M.tb were considered from the target library of Molecular Property Diagnostic Suite-Tuberculosis (MPDS^TB- http://mpds.neist.res.in:8084 ) for virtual screening. FDA approved drugs were collected, preprocessed and docked in the active sites of the 20 M.tb targets. The top 300 drug molecules from each target (20 × 300) were filtered-in and subsequently screened for possible antitubercular and antimycobacterial activity using PASS tool. Using this approach, 34 drugs with predicted antitubercular and anti-mycobacterial activity were identified along with good binding affinity against multiple M.tb targets. Interestingly, 21 out of the 34 identified drugs are antibiotics while 4 drug molecules (nitrofural, stavudine, quinine and quinidine) are non-antibiotics showing promising predicted antitubercular activity. Most of these molecules have the similar privileged antimycobacterial drugs scaffold. Further drug likeness properties were calculated to get deeper insights to M.tb lead molecules. Interestingly, it was also observed that the drugs identified from the study are under different stages of drug discovery (i.e., in vitro, clinical trials) for the effective treatment of various diseases including cancer, degenerative diseases, dengue virus infection, tuberculosis, etc. Krasavin et al., 2017 synthesized nitrofuran analogues with appreciable MICs (22-23 µM) against M.tb H37Rv. These experiments further add to the credibility of the drugs identified in this study (TB).