Predicting phospholipidosis using machine learning

Construction of Discrimination Models of Cationic Drugs for Phospholipidosis Induction Potential by Using Interaction Data with Immobilized Artificial Membrane as Well as Physicochemical Properties

Accurate prediction of the phospholipidosis-induction risk of drugs at early stages is important in drug development. So far, discrimination models for predicting the induction risk of cationic drugs have been proposed, but it is still challenging to accurately predict the risk of cationic drugs with intermediate hydrophobicity (logP). In this study, we introduced a parameter (Δlogk40) reflecting not only hydrophobic interaction but also interactions with the polar headgroup between cationic drugs and phospholipids, obtained with liquid chromatography using an immobilized artificial membrane column. The parameter was used along with other physicochemical properties as features to construct discrimination models. Linear discriminant analysis, the modified Mahalanobis discriminant analysis, support vector machine, and random forest were employed for model construction. The results showed that all discrimination models exhibited good predictive performance, with the modified Mahalanobis discriminant analysis and random forest providing the best results for cationic drugs, suggesting that the usefulness of the parameter reflecting complex interactions between cationic drugs and immobilized artificial membrane for constructing discrimination models to predict the induction risk. Furthermore, by applying the parameter as a feature in constructing discrimination models, we demonstrated an improvement in the predictive performance for drugs with intermediate hydrophobicity.

AMALPHI: A Machine Learning Platform for Predicting Drug-Induced PhospholIpidosis

Drug-induced phospholipidosis (PLD) involves the accumulation of phospholipids in cells of multiple tissues, particularly within lysosomes, and it is associated with prolonged exposure to druglike compounds, predominantly cationic amphiphilic drugs (CADs). PLD affects a significant portion of drugs currently in development and has recently been proven to be responsible for confounding antiviral data during drug repurposing for SARS-CoV-2. In these scenarios, it has become crucial to identify potential safe drug candidates in advance and distinguish them from those that may lead to false in vitro antiviral activity. In this work, we developed a series of machine learning classifiers with the aim of predicting the PLD-inducing potential of drug candidates. The models were built on a high-quality chemical collection comprising 545 curated small molecules extracted from ChEMBL v30. The most effective model, obtained using the balanced random forest algorithm, achieved high performance, including an AUC value computed in validation as high as 0.90. The model was made freely available through a user-friendly web platform named AMALPHI (https://www.ba.ic.cnr.it/softwareic/amalphiportal/), which can represent a valuable tool for medicinal chemists interested in conducting an early evaluation of PLD inducer potential.

A machine learning and live-cell imaging tool kit uncovers small molecules induced phospholipidosis

Drug-induced phospholipidosis (DIPL), characterized by excessive accumulation of phospholipids in lysosomes, can lead to clinical adverse effects. It may also alter phenotypic responses in functional studies using chemical probes. Therefore, robust methods are needed to predict and quantify phospholipidosis (PL) early in drug discovery and in chemical probe characterization. Here, we present a versatile high-content live-cell imaging approach, which was used to evaluate a chemogenomic and a lysosomal modulation library. We trained and evaluated several machine learning models using the most comprehensive set of publicly available compounds and interpreted the best model using SHapley Additive exPlanations (SHAP). Analysis of high-quality chemical probes extracted from the Chemical Probes Portal using our algorithm revealed that closely related molecules, such as chemical probes and their matched negative controls can differ in their ability to induce PL, highlighting the importance of identifying PL for robust target validation in chemical biology.

Drug-induced phospholipidosis confounds drug repurposing for SARS-CoV-2

Repurposing drugs as treatments for COVID-19, the disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has drawn much attention. Beginning with sigma receptor ligands and expanding to other drugs from screening in the field, we became concerned that phospholipidosis was a shared mechanism underlying the antiviral activity of many repurposed drugs. For all of the 23 cationic amphiphilic drugs we tested, including hydroxychloroquine, azithromycin, amiodarone, and four others already in clinical trials, phospholipidosis was monotonically correlated with antiviral efficacy. Conversely, drugs active against the same targets that did not induce phospholipidosis were not antiviral. Phospholipidosis depends on the physicochemical properties of drugs and does not reflect specific target-based activities-rather, it may be considered a toxic confound in early drug discovery. Early detection of phospholipidosis could eliminate these artifacts, enabling a focus on molecules with therapeutic potential.

Phospholipidosis is a shared mechanism underlying the in vitro antiviral activity of many repurposed drugs against SARS-CoV-2

Repurposing drugs as treatments for COVID-19 has drawn much attention. A common strategy has been to screen for established drugs, typically developed for other indications, that are antiviral in cells or organisms. Intriguingly, most of the drugs that have emerged from these campaigns, though diverse in structure, share a common physical property: cationic amphiphilicity. Provoked by the similarity of these repurposed drugs to those inducing phospholipidosis, a well-known drug side effect, we investigated phospholipidosis as a mechanism for antiviral activity. We tested 23 cationic amphiphilic drugs-including those from phenotypic screens and others that we ourselves had found-for induction of phospholipidosis in cell culture. We found that most of the repurposed drugs, which included hydroxychloroquine, azithromycin, amiodarone, and four others that have already progressed to clinical trials, induced phospholipidosis in the same concentration range as their antiviral activity; indeed, there was a strong monotonic correlation between antiviral efficacy and the magnitude of the phospholipidosis. Conversely, drugs active against the same targets that did not induce phospholipidosis were not antiviral. Phospholipidosis depends on the gross physical properties of drugs, and does not reflect specific target-based activities, rather it may be considered a confound in early drug discovery. Understanding its role in infection, and detecting its effects rapidly, will allow the community to better distinguish between drugs and lead compounds that more directly impact COVID-19 from the large proportion of molecules that manifest this confounding effect, saving much time, effort and cost.

Repurposing therapeutics for COVID-19: Rapid prediction of commercially available drugs through machine learning and docking

BACKGROUND

The outbreak of the novel coronavirus disease COVID-19, caused by the SARS-CoV-2 virus has spread rapidly around the globe during the past 3 months. As the virus infected cases and mortality rate of this disease is increasing exponentially, scientists and researchers all over the world are relentlessly working to understand this new virus along with possible treatment regimens by discovering active therapeutic agents and vaccines. So, there is an urgent requirement of new and effective medications that can treat the disease caused by SARS-CoV-2.

METHODS AND FINDINGS

We perform the study of drugs that are already available in the market and being used for other diseases to accelerate clinical recovery, in other words repurposing of existing drugs. The vast complexity in drug design and protocols regarding clinical trials often prohibit developing various new drug combinations for this epidemic disease in a limited time. Recently, remarkable improvements in computational power coupled with advancements in Machine Learning (ML) technology have been utilized to revolutionize the drug development process. Consequently, a detailed study using ML for the repurposing of therapeutic agents is urgently required. Here, we report the ML model based on the Naive Bayes algorithm, which has an accuracy of around 73% to predict the drugs that could be used for the treatment of COVID-19. Our study predicts around ten FDA approved commercial drugs that can be used for repurposing. Among all, we found that 3 of the drugs fulfils the criterions well among which the antiretroviral drug Amprenavir (DrugBank ID-DB00701) would probably be the most effective drug based on the selected criterions.

CONCLUSIONS

Our study can help clinical scientists in being more selective in identifying and testing the therapeutic agents for COVID-19 treatment. The ML based approach for drug discovery as reported here can be a futuristic smart drug designing strategy for community applications.

Comparison of Machine Learning Models on Performance of Single- and Dual-Type Electrochromic Devices

This study shows that the model fitting based on machine learning (ML) from experimental data can successfully predict the electrochromic characteristics of single- and dual-type flexible electrochromic devices (ECDs) by using tungsten trioxide (WO3) and WO3/vanadium pentoxide (V2O5), respectively. Seven different regression methods were used for experimental observations, which belong to single and dual ECDs where 80% percent was used as training data and the remaining was taken as testing data. Among the seven different regression methods, K-nearest neighbor (KNN) achieves the best results with higher coefficient of determination (R 2) score and lower root-mean-squared error (RMSE) for the bleaching state of ECDs. Furthermore, higher R 2 score and lower RMSE for the coloration state of ECDs were achieved with Gaussian process regressor. The robustness result of the ML modeling demonstrates the reliability of prediction outcomes. These results can be proposed as promising models for different energy-saving flexible electronic systems.

Qualified method for the estimation of di-18:1 bis(monoacylglycero)phosphate in urine, a noninvasive biomarker to monitor drug-induced phospholipidosis

Aim: Our objective was to develop and qualify a bioanalytical method for the estimation of di-18:1-bis(monoacylglycero)phosphate (di-18:1 BMP) as a urinary biomarker for the assessment of drug-induced phospholipidosis and demonstrate its application in a preclinical study. Methodology/results: di-18:1 BMP was extracted by liquid-liquid extraction using n-butanol and analyzed by LC-MS/MS. The qualified method was selective, precise, robust and accurate across the linearity range (0.2-250 ng/ml). Qualified method was then used to assess chloroquine-induced phospholipidosis in rats dosed at 120 mg/kg for 5 days. A fivefold increase in di-18:1 BMP was observed on Day 5 compared with predose. Conclusion: Di-18:1 BMP can be used as a noninvasive biomarker to assess/screen compounds that could cause drug-induced phospholipidosis in rats.

Establishing GPCR Targets of hMAO Active Anthraquinones from Cassia obtusifolia Linn Seeds Using In Silico and In Vitro Methods

The present study examines the effect of human monoamine oxidase active anthraquinones emodin, alaternin (=7-hydroxyemodin), aloe-emodin, and questin from Cassia obtusifolia Linn seeds in modulating human dopamine (hD₁R, hD₃R, and hD₄R), serotonin (h5-HT_1AR), and vasopressin (hV_1AR) receptors that were predicted as prime targets from proteocheminformatics modeling via in vitro cell-based functional assays, and explores the possible mechanisms of action via in silico modeling. Emodin and alaternin showed a concentration-dependent agonist effect on hD₃R with EC₅₀ values of 21.85 ± 2.66 and 56.85 ± 4.59 μM, respectively. On hV_1AR, emodin and alaternin showed an antagonist effect with IC₅₀ values of 10.25 ± 1.97 and 11.51 ± 1.08 μM, respectively. Interestingly, questin and aloe-emodin did not have any observable effect on hV_1AR. Only alaternin was effective in antagonizing h5-HT_1AR (IC₅₀: 84.23 ± 4.12 μM). In silico studies revealed that a hydroxyl group at C1, C3, and C8 and a methyl group at C6 of anthraquinone structure are essential for hD₃R agonist and hV_1AR antagonist effects, as well as for the H-bond interaction of 1-OH group with Ser192 at a proximity of 2.0 Å. Thus, based on in silico and in vitro results, hV_1AR, hD₃R, and h5-HT_1AR appear to be prime targets of the tested anthraquinones.

Exploiting ensemble learning to improve prediction of phospholipidosis inducing potential

Phospholipidosis is characterized by the presence of excessive accumulation of phospholipids in different tissue types (lungs, liver, eyes, kidneys etc.) caused by cationic amphiphilic drugs. Electron microscopy analysis has revealed the presence of lamellar inclusion bodies as the hallmark of phospholipidosis. Some phospholipidosis causing compounds can cause tissue specific inflammatory/retrogressive changes. Reliable and accurate in silico methods could facilitate early screening of phospholipidosis inducing compounds which can subsequently speed up the pharmaceutical drug discovery pipelines. In the present work, stacking ensembles are implemented for combining a number of different base learners to develop predictive models (a total of 256 trained machine learning models were tested) for phospholipidosis inducing compounds using a wide range of molecular descriptors (ChemMine, JOELib, Open babel and RDK descriptors) and structural alerts as input features. The best model consisting of stacked ensemble of machine learning algorithms with random forest as the second level learner outperformed other base and ensemble learners. JOELib descriptors along with structural alerts performed better than the other types of descriptor sets. The best ensemble model achieved an overall accuracy of 88.23%, sensitivity of 86.27%, specificity of 90.20%, mcc of 0.765, auc of 0.896 with 88.21 g-means. To assess the robustness and stability of the best ensemble model, it is further evaluated using stratified 10×10 fold cross validation and holdout testing sets (repeated 10 times) achieving 84.83% mean accuracy with 0.708 mean mcc and 88.46% mean accuracy with 0.771 mean mcc respectively. A comparison of different meta classifiers (Generalized linear regression, Gradient boosting machines, Random forest and Deep learning neural networks) in stacking ensemble revealed that random forest is the better choice for combining multiple classification models.

Characterizing Adversity of Lysosomal Accumulation in Nonclinical Toxicity Studies: Results from the 5th ESTP International Expert Workshop

Lysosomes have a central role in cellular catabolism, trafficking, and processing of foreign particles. Accumulation of endogenous and exogenous materials in lysosomes represents a common finding in nonclinical toxicity studies. Histologically, these accumulations often lack distinctive features indicative of lysosomal or cellular dysfunction, making it difficult to consistently interpret and assign adverse dose levels. To help address this issue, the European Society of Toxicologic Pathology organized a workshop where representative types of lysosomal accumulation induced by pharmaceuticals and environmental chemicals were presented and discussed. The expert working group agreed that the diversity of lysosomal accumulations requires a case-by-case weight-of-evidence approach and outlined several factors to consider in the adversity assessment, including location and type of cell affected, lysosomal contents, severity of the accumulation, and related pathological effects as evidence of cellular or organ dysfunction. Lysosomal accumulations associated with cytotoxicity, inflammation, or fibrosis were generally considered to be adverse, while those found in isolation (without morphologic or functional consequences) were not. Workshop examples highlighted the importance of thoroughly characterizing the biological context of lysosomal effects, including mechanistic data and functional in vitro readouts if available. The information provided here should facilitate greater consistency and transparency in the interpretation of lysosomal effects.

Derivatization of common antidepressant drugs increases inhibition of acid sphingomyelinase and reduces induction of phospholipidosis

In recent studies, major depressive disorder (MDD) was linked to an increase in acid sphingomyelinase (ASM) activity. Several drugs that are commonly used to treat MDD functionally inhibit the lysosomal enzyme ASM and are called functional inhibitors of ASM (FIASMAs). These drugs are classified as cationic amphiphilic drugs (CADs) that influence the catalytic activities of different lysosomal enzymes. This action results in the side effect of phospholipidosis (PLD), which describes a detrimental increase in the phospholipid content in lysosomes. FIASMAs differ only slightly in their physico-chemical properties, but their effects on ASM activity and induction of the lysosomal phospholipid content vary significantly. In this study, we systematically induced minor chemical modifications to the FIASMAs imipramine, desipramine and fluoxetine. We generated a library of 45 new CADs with slightly different log P (logarithmic partition coefficient) and pKa (logarithmic acid dissociation constant) values. The effects of the compounds on the ASM activity and lysosomal phospholipid content were assessed in cell culture assays. We identified four compounds with beneficial effects, i.e., increased ASM activity inhibition and reduced PLD induction compared with the original drugs. The compounds HT04, RH272B and RH272D outperformed the original imipramine, whereas RH281A performed better than desipramine. Thus, minor chemical variations of CADs impact lysosomal metabolism in a specific manner and can lead to antidepressant drugs with less deleterious side effects.

QSAR based predictive modeling for anti-malarial molecules

Malaria is a predominant infectious disease, with a global footprint, but especially severe in developing countries in the African subcontinent. In recent years, drug-resistant malaria has become an alarming factor, and hence the requirement of new and improved drugs is more crucial than ever before. One of the promising locations for antimalarial drug target is the apicoplast, as this organelle does not occur in humans. The apicoplast is associated with many unique and essential pathways in many Apicomplexan pathogens, including Plasmodium. The use of machine learning methods is now commonly available through open source programs. In the present work, we describe a standard protocol to develop molecular descriptor based predictive models (QSAR models), which can be further utilized for the screening of large chemical libraries. This protocol is used to build models using training data sourced from apicoplast specific bioassays. Multiple model building methods are used including Generalized Linear Models (GLM), Random Forest (RF), C5.0 implementation of a decision tree, Support Vector Machines (SVM), K-Nearest Neighbour and Naive Bayes. Methods to evaluate the accuracy of the model building method are included in the protocol. For the given dataset, the C5.0, SVM and RF perform better than other methods, with comparable accuracy over the test data.

Machine-learning approaches in drug discovery: methods and applications

During the past decade, virtual screening (VS) has evolved from traditional similarity searching, which utilizes single reference compounds, into an advanced application domain for data mining and machine-learning approaches, which require large and representative training-set compounds to learn robust decision rules. The explosive growth in the amount of public domain-available chemical and biological data has generated huge effort to design, analyze, and apply novel learning methodologies. Here, I focus on machine-learning techniques within the context of ligand-based VS (LBVS). In addition, I analyze several relevant VS studies from recent publications, providing a detailed view of the current state-of-the-art in this field and highlighting not only the problematic issues, but also the successes and opportunities for further advances.

Enzyme informatics

Over the last 50 years, sequencing, structural biology and bioinformatics have completely revolutionised biomolecular science, with millions of sequences and tens of thousands of three dimensional structures becoming available. The bioinformatics of enzymes is well served by, mostly free, online databases. BRENDA describes the chemistry, substrate specificity, kinetics, preparation and biological sources of enzymes, while KEGG is valuable for understanding enzymes and metabolic pathways. EzCatDB, SFLD and MACiE are key repositories for data on the chemical mechanisms by which enzymes operate. At the current rate of genome sequencing and manual annotation, human curation will never finish the functional annotation of the ever-expanding list of known enzymes. Hence there is an increasing need for automated annotation, though it is not yet widespread for enzyme data. In contrast, functional ontologies such as the Gene Ontology already profit from automation. Despite our growing understanding of enzyme structure and dynamics, we are only beginning to be able to design novel enzymes. One can now begin to trace the functional evolution of enzymes using phylogenetics. The ability of enzymes to perform secondary functions, albeit relatively inefficiently, gives clues as to how enzyme function evolves. Substrate promiscuity in enzymes is one example of imperfect specificity in protein-ligand interactions. Similarly, most drugs bind to more than one protein target. This may sometimes result in helpful polypharmacology as a drug modulates plural targets, but also often leads to adverse side-effects. Many chemoinformatics approaches can be used to model the interactions between druglike molecules and proteins in silico. We can even use quantum chemical techniques like DFT and QM/MM to compute the structural and energetic course of enzyme catalysed chemical reaction mechanisms, including a full description of bond making and breaking.

Machine learning methods in chemoinformatics

Machine learning algorithms are generally developed in computer science or adjacent disciplines and find their way into chemical modeling by a process of diffusion. Though particular machine learning methods are popular in chemoinformatics and quantitative structure-activity relationships (QSAR), many others exist in the technical literature. This discussion is methods-based and focused on some algorithms that chemoinformatics researchers frequently use. It makes no claim to be exhaustive. We concentrate on methods for supervised learning, predicting the unknown property values of a test set of instances, usually molecules, based on the known values for a training set. Particularly relevant approaches include Artificial Neural Networks, Random Forest, Support Vector Machine, k-Nearest Neighbors and naïve Bayes classifiers.

Role of electrostatic potential in the in silico prediction of molecular bioactivation and mutagenesis

Electrostatic potential (ESP) is a useful physicochemical property of a molecule that provides insights into inter- and intramolecular associations, as well as prediction of likely sites of electrophilic and nucleophilic metabolic attack. Knowledge of sites of metabolic attack is of paramount importance in DMPK research since drugs frequently fail in clinical trials due to the formation of bioactivated metabolites which are often difficult to measure experimentally due to their reactive nature and relatively short half-lives. Computational chemistry methods have proven invaluable in recent years as a means to predict and study bioactivated metabolites without the need for chemical syntheses, or testing on experimental animals. Additional molecular properties (heat of formation, heat of solvation and E(LUMO) - E(HOMO)) are discussed in this paper as complementary indicators of the behavior of metabolites in vivo. Five diverse examples are presented (acetaminophen, aniline/phenylamines, imidacloprid, nefazodone and vinyl chloride) which illustrate the utility of this multidimensional approach in predicting bioactivation, and in each case the predicted data agreed with experimental data described in the scientific literature. A further example of the usefulness of calculating ESP, in combination with the molecular properties mentioned above, is provided by an examination of the use of these parameters in providing an explanation for the sites of nucleophilic attack of the nucleic acid cytosine. Exploration of sites of nucleophilic attack of nucleic acids is important as adducts of DNA have the potential to result in mutagenesis.

Identification of drugs inducing phospholipidosis by novel in vitro data

Drug-induced phospholipidosis (PLD) is a lysosomal storage disorder characterized by the accumulation of phospholipids within the lysosome. This adverse drug effect can occur in various tissues and is suspected to impact cellular viability. Therefore, it is important to test chemical compounds for their potential to induce PLD during the drug design process. PLD has been reported to be a side effect of many commonly used drugs, especially those with cationic amphiphilic properties. To predict drug-induced PLD in silico, we established a high-throughput cell-culture-based method to quantitatively determine the induction of PLD by chemical compounds. Using this assay, we tested 297 drug-like compounds at two different concentrations (2.5 μM and 5.0 μM). We were able to identify 28 previously unknown PLD-inducing agents. Furthermore, our experimental results enabled the development of a binary classification model to predict PLD-inducing agents based on their molecular properties. This random forest prediction system yields a bootstrapped validated accuracy of 86 %. PLD-inducing agents overlap with those that target similar biological processes; a high degree of concordance with PLD-inducing agents was identified for cationic amphiphilic compounds, small molecules that inhibit acid sphingomyelinase, compounds that cross the blood-brain barrier, and compounds that violate Lipinski's rule of five. Furthermore, we were able to show that PLD-inducing compounds applied in combination additively induce PLD.

Drug induced phospholipidosis: an acquired lysosomal storage disorder

There is a strong association between lysosome enzyme deficiencies and monogenic disorders resulting in lysosomal storage disease. Of the more than 75 characterized lysosomal proteins, two thirds are directly linked to inherited diseases of metabolism. Only one lysosomal storage disease, Niemann-Pick disease, is associated with impaired phospholipid metabolism. However, other phospholipases are found in the lysosome but remain poorly characterized. A recent exception is lysosomal phospholipase A2 (group XV phospholipase A2). Although no inherited disorder of lysosomal phospholipid metabolism has yet been associated with a loss of function of this lipase, this enzyme may be a target for an acquired form of lysosomal storage, drug induced phospholipidosis. This article is part of a Special Issue entitled Phospholipids and Phospholipid Metabolism.

Is EC class predictable from reaction mechanism?

BACKGROUND

We investigate the relationships between the EC (Enzyme Commission) class, the associated chemical reaction, and the reaction mechanism by building predictive models using Support Vector Machine (SVM), Random Forest (RF) and k-Nearest Neighbours (kNN). We consider two ways of encoding the reaction mechanism in descriptors, and also three approaches that encode only the overall chemical reaction. Both cross-validation and also an external test set are used.

RESULTS

The three descriptor sets encoding overall chemical transformation perform better than the two descriptions of mechanism. SVM and RF models perform comparably well; kNN is less successful. Oxidoreductases and hydrolases are relatively well predicted by all types of descriptor; isomerases are well predicted by overall reaction descriptors but not by mechanistic ones.

CONCLUSIONS

Our results suggest that pairs of similar enzyme reactions tend to proceed by different mechanisms. Oxidoreductases, hydrolases, and to some extent isomerases and ligases, have clear chemical signatures, making them easier to predict than transferases and lyases. We find evidence that isomerases as a class are notably mechanistically diverse and that their one shared property, of substrate and product being isomers, can arise in various unrelated ways.The performance of the different machine learning algorithms is in line with many cheminformatics applications, with SVM and RF being roughly equally effective. kNN is less successful, given the role that non-local information plays in successful classification. We note also that, despite a lack of clarity in the literature, EC number prediction is not a single problem; the challenge of predicting protein function from available sequence data is quite different from assigning an EC classification from a cheminformatics representation of a reaction.

Computational models for in-vitro anti-tubercular activity of molecules based on high-throughput chemical biology screening datasets

Background

The emergence of Multi-drug resistant tuberculosis in pandemic proportions throughout the world and the paucity of novel therapeutics for tuberculosis have re-iterated the need to accelerate the discovery of novel molecules with anti-tubercular activity. Though high-throughput screens for anti-tubercular activity are available, they are expensive, tedious and time-consuming to be performed on large scales. Thus, there remains an unmet need to prioritize the molecules that are taken up for biological screens to save on cost and time. Computational methods including Machine Learning have been widely employed to build classifiers for high-throughput virtual screens to prioritize molecules for further analysis. The availability of datasets based on high-throughput biological screens or assays in public domain makes computational methods a plausible proposition for building predictive models. In addition, this approach would save significantly on the cost, effort and time required to run high throughput screens.

Results

We show that by using four supervised state-of-the-art classifiers (SMO, Random Forest, Naive Bayes and J48) we are able to generate in-silico predictive models on an extremely imbalanced (minority class ratio: 0.6%) large dataset of anti-tubercular molecules with reasonable AROC (0.6-0.75) and BCR (60-66%) values. Moreover, these models are able to provide 3-4 fold enrichment over random selection.

Conclusions

In the present study, we have used the data from in-vitro screens for anti-tubercular activity from a high-throughput screen available in public domain to build highly accurate classifiers based on molecular descriptors of the molecules. We show that Machine Learning tools can be used to build highly effective predictive models for virtual high-throughput screens to prioritize molecules from large molecular libraries.

Predicting the mechanism of phospholipidosis

The mechanism of phospholipidosis is still not well understood. Numerous different mechanisms have been proposed, varying from direct inhibition of the breakdown of phospholipids to the binding of a drug compound to the phospholipid, preventing breakdown. We have used a probabilistic method, the Parzen-Rosenblatt Window approach, to build a model from the ChEMBL dataset which can predict from a compound's structure both its primary pharmaceutical target and other targets with which it forms off-target, usually weaker, interactions. Using a small dataset of 182 phospholipidosis-inducing and non-inducing compounds, we predict their off-target activity against targets which could relate to phospholipidosis as a side-effect of a drug. We link these targets to specific mechanisms of inducing this lysosomal build-up of phospholipids in cells. Thus, we show that the induction of phospholipidosis is likely to occur by separate mechanisms when triggered by different cationic amphiphilic drugs. We find that both inhibition of phospholipase activity and enhanced cholesterol biosynthesis are likely to be important mechanisms. Furthermore, we provide evidence suggesting four specific protein targets. Sphingomyelin phosphodiesterase, phospholipase A2 and lysosomal phospholipase A1 are shown to be likely targets for the induction of phospholipidosis by inhibition of phospholipase activity, while lanosterol synthase is predicted to be associated with phospholipidosis being induced by enhanced cholesterol biosynthesis. This analysis provides the impetus for further experimental tests of these hypotheses.

Predictive models for anti-tubercular molecules using machine learning on high-throughput biological screening datasets

Background

Tuberculosis is a contagious disease caused by Mycobacterium tuberculosis (Mtb), affecting more than two billion people around the globe and is one of the major causes of morbidity and mortality in the developing world. Recent reports suggest that Mtb has been developing resistance to the widely used anti-tubercular drugs resulting in the emergence and spread of multi drug-resistant (MDR) and extensively drug-resistant (XDR) strains throughout the world. In view of this global epidemic, there is an urgent need to facilitate fast and efficient lead identification methodologies. Target based screening of large compound libraries has been widely used as a fast and efficient approach for lead identification, but is restricted by the knowledge about the target structure. Whole organism screens on the other hand are target-agnostic and have been now widely employed as an alternative for lead identification but they are limited by the time and cost involved in running the screens for large compound libraries. This could be possibly be circumvented by using computational approaches to prioritize molecules for screening programmes.

Results

We utilized physicochemical properties of compounds to train four supervised classifiers (Naïve Bayes, Random Forest, J48 and SMO) on three publicly available bioassay screens of Mtb inhibitors and validated the robustness of the predictive models using various statistical measures.

Conclusions

This study is a comprehensive analysis of high-throughput bioassay data for anti-tubercular activity and the application of machine learning approaches to create target-agnostic predictive models for anti-tubercular agents.

Informatics, machine learning and computational medicinal chemistry

This article reviews the use of informatics and computational chemistry methods in medicinal chemistry, with special consideration of how computational techniques can be adapted and extended to obtain more and higher-quality information. Special consideration is given to the computation of protein–ligand binding affinities, to the prediction of off-target bioactivities, bioactivity spectra and computational toxicology, and also to calculating absorption-, distribution-, metabolism- and excretion-relevant properties, such as solubility.