High throughput screening of a library based on kinase inhibitor scaffolds against Mycobacterium tuberculosis H37Rv

Are bigger data sets better for machine learning? Fusing single-point and dual-event dose response data for Mycobacterium tuberculosis

Tuberculosis is a major, neglected disease for which the quest to find new treatments continues. There is an abundance of data from large phenotypic screens in the public domain against Mycobacterium tuberculosis (Mtb). Since machine learning methods can learn from past data, we were interested in addressing whether more data builds better models. We now describe using Bayesian machine learning to assess whether we can improve our models by combining the large quantities of single-point data with the much smaller (higher quality) dual-event data sets, which use both dose-response data for both whole-cell antitubercular activity and Vero cell cytotoxicity. We have evaluated 12 models ranging from different single-point, dual-event dose-response, single-point and dual-event dose-response as well as combined data sets for three distinct data sets from the same laboratory. We used a fourth data set of active and inactive compounds from the same group as well as a smaller set of 177 active compounds from GlaxoSmithKline as test sets. Our data suggest combining single-point with dual-event dose-response data does not diminish the internal or external predictive ability of the models based on the receiver operator curve (ROC) for these models (internal ROC range 0.83-0.91, external ROC range 0.62-0.83) compared to the orders of magnitude smaller dual-event models (internal ROC range 0.6-0.83 and external ROC 0.54-0.83). In conclusion, models developed with 1200-5000 compounds appear to be as predictive as those generated with 25 000-350 000 molecules. Our results have implications for justifying further high-throughput screening versus focused testing based on model predictions.

Pros and cons of the tuberculosis drugome approach--an empirical analysis

Drug-resistant Mycobacterium tuberculosis (MTB), the causative pathogen of tuberculosis (TB), has become a serious threat to global public health. Yet the development of novel drugs against MTB has been lagging. One potentially powerful approach to drug development is computation-aided repositioning of current drugs. However, the effectiveness of this approach has rarely been examined. Here we select the "TB drugome" approach--a protein structure-based method for drug repositioning for tuberculosis treatment--to (1) experimentally validate the efficacy of the identified drug candidates for inhibiting MTB growth, and (2) computationally examine how consistently drug candidates are prioritized, considering changes in input data. Twenty three drugs in the TB drugome were tested. Of them, only two drugs (tamoxifen and 4-hydroxytamoxifen) effectively suppressed MTB growth at relatively high concentrations. Both drugs significantly enhanced the inhibitory effects of three first-line anti-TB drugs (rifampin, isoniazid, and ethambutol). However, tamoxifen is not a top-listed drug in the TB drugome, and 4-hydroxytamoxifen is not approved for use in humans. Computational re-examination of the TB drugome indicated that the rankings were subject to technical and data-related biases. Thus, although our results support the effectiveness of the TB drugome approach for identifying drugs that can potentially be repositioned for stand-alone applications or for combination treatments for TB, the approach requires further refinements via incorporation of additional biological information. Our findings can also be extended to other structure-based drug repositioning methods.

Looking back to the future: predicting in vivo efficacy of small molecules versus Mycobacterium tuberculosis

Selecting and translating in vitro leads for a disease into molecules with in vivo activity in an animal model of the disease is a challenge that takes considerable time and money. As an example, recent years have seen whole-cell phenotypic screens of millions of compounds yielding over 1500 inhibitors of Mycobacterium tuberculosis (Mtb). These must be prioritized for testing in the mouse in vivo assay for Mtb infection, a validated model utilized to select compounds for further testing. We demonstrate learning from in vivo active and inactive compounds using machine learning classification models (Bayesian, support vector machines, and recursive partitioning) consisting of 773 compounds. The Bayesian model predicted 8 out of 11 additional in vivo actives not included in the model as an external test set. Curation of 70 years of Mtb data can therefore provide statistically robust computational models to focus resources on in vivo active small molecule antituberculars. This highlights a cost-effective predictor for in vivo testing elsewhere in other diseases.

Bayesian models for screening and TB Mobile for target inference with Mycobacterium tuberculosis

The search for compounds active against Mycobacterium tuberculosis is reliant upon high-throughput screening (HTS) in whole cells. We have used Bayesian machine learning models which can predict anti-tubercular activity to filter an internal library of over 150,000 compounds prior to in vitro testing. We used this to select and test 48 compounds in vitro; 11 were active with MIC values ranging from 0.4 μM to 10.2 μM, giving a high hit rate of 22.9%. Among the hits, we identified several compounds belonging to the same series including five quinolones (including ciprofloxacin), three molecules with long aliphatic linkers and three singletons. This approach represents a rapid method to prioritize compounds for testing that can be used alongside medicinal chemistry insight and other filters to identify active molecules. Such models can significantly increase the hit rate of HTS, above the usual 1% or lower rates seen. In addition, the potential targets for the 11 molecules were predicted using TB Mobile and clustering alongside a set of over 740 molecules with known M. tuberculosis target annotations. These predictions may serve as a mechanism for prioritizing compounds for further optimization.

Development of novel tetrahydrothieno[2,3-c]pyridine-3-carboxamide based Mycobacterium tuberculosis pantothenate synthetase inhibitors: molecular hybridization from known antimycobacterial leads

Twenty six 2,6-disubstituted 4,5,6,7-tetrahydrothieno[2,3-c]pyridine-3-carboxamide derivatives were designed by molecular hybridization approach using and synthesized from piperidin-4-one by five step synthesis. Compounds were evaluated for Mycobacterium tuberculosis (MTB) pantothenate synthetase (PS) inhibition study, in vitro activities against MTB, cytotoxicity against RAW 264.7 cell line. Among the compounds, 6-(4-nitrophenylsulfonyl)-2-(5-nitrothiophene-2-carboxamido)-4,5,6,7-tetrahydrothieno[2,3-c]pyridine-3-carboxamide (11) was found to be the most active compound with IC50 of 5.87 ± 0.12 μM against MTB PS, inhibited MTB with MIC of 9.28 μM and it was non-cytotoxic at 50 μM. The binding affinity of the most potent inhibitor 11 was further confirmed biophysically through differential scanning fluorimetry.

Fusing dual-event data sets for Mycobacterium tuberculosis machine learning models and their evaluation

The search for new tuberculosis treatments continues as we need to find molecules that can act more quickly, be accommodated in multidrug regimens, and overcome ever increasing levels of drug resistance. Multiple large scale phenotypic high-throughput screens against Mycobacterium tuberculosis (Mtb) have generated dose response data, enabling the generation of machine learning models. These models also incorporated cytotoxicity data and were recently validated with a large external data set. A cheminformatics data-fusion approach followed by Bayesian machine learning, Support Vector Machine, or Recursive Partitioning model development (based on publicly available Mtb screening data) was used to compare individual data sets and subsequent combined models. A set of 1924 commercially available molecules with promising antitubercular activity (and lack of relative cytotoxicity to Vero cells) were used to evaluate the predictive nature of the models. We demonstrate that combining three data sets incorporating antitubercular and cytotoxicity data in Vero cells from our previous screens results in external validation receiver operator curve (ROC) of 0.83 (Bayesian or RP Forest). Models that do not have the highest 5-fold cross-validation ROC scores can outperform other models in a test set dependent manner. We demonstrate with predictions for a recently published set of Mtb leads from GlaxoSmithKline that no single machine learning model may be enough to identify compounds of interest. Data set fusion represents a further useful strategy for machine learning construction as illustrated with Mtb. Coverage of chemistry and Mtb target spaces may also be limiting factors for the whole-cell screening data generated to date.

Curing TB with open science

There are many funded efforts going on focused on tuberculosis research and drug or vaccine development. There is little if any global coordination or collaboration and subsequently there are likely to be huge data silos and duplication of efforts. We now propose steps to remedy this by fostering more open science in TB research.

Identification of novel inhibitors of nonreplicating Mycobacterium tuberculosis using a carbon starvation model

During Mycobacterium tuberculosis infection, a population of bacteria is thought to exist in a nonreplicating state, refractory to antibiotics, which may contribute to the need for prolonged antibiotic therapy. The identification of inhibitors of the nonreplicating state provides tools that can be used to probe this hypothesis and the physiology of this state. The development of such inhibitors also has the potential to shorten the duration of antibiotic therapy required. Here we describe the development of a novel nonreplicating assay amenable to high-throughput chemical screening coupled with secondary assays that use carbon starvation as the in vitro model. Together these assays identify compounds with activity against replicating and nonreplicating M. tuberculosis as well as compounds that inhibit the transition from nonreplicating to replicating stages of growth. Using these assays we successfully screened over 300,000 compounds and identified 786 inhibitors of nonreplicating M. tuberculosis In order to understand the relationship among different nonreplicating models, we tested 52 of these molecules in a hypoxia model, and four different chemical scaffolds in a stochastic persister model, and a streptomycin-dependent model. We found that compounds display varying levels of activity in different models for the nonreplicating state, suggesting important differences in bacterial physiology between models. Therefore, chemical tools identified in this assay may be useful for determining the relevance of different nonreplicating in vitro models to in vivo M. tuberculosis infection. Given our current limited understanding, molecules that are active across multiple models may represent more promising candidates for further development.

Combining computational methods for hit to lead optimization in Mycobacterium tuberculosis drug discovery

PURPOSE

Tuberculosis treatments need to be shorter and overcome drug resistance. Our previous large scale phenotypic high-throughput screening against Mycobacterium tuberculosis (Mtb) has identified 737 active compounds and thousands that are inactive. We have used this data for building computational models as an approach to minimize the number of compounds tested.

METHODS

A cheminformatics clustering approach followed by Bayesian machine learning models (based on publicly available Mtb screening data) was used to illustrate that application of these models for screening set selections can enrich the hit rate.

RESULTS

In order to explore chemical diversity around active cluster scaffolds of the dose-response hits obtained from our previous Mtb screens a set of 1924 commercially available molecules have been selected and evaluated for antitubercular activity and cytotoxicity using Vero, THP-1 and HepG2 cell lines with 4.3%, 4.2% and 2.7% hit rates, respectively. We demonstrate that models incorporating antitubercular and cytotoxicity data in Vero cells can significantly enrich the selection of non-toxic actives compared to random selection. Across all cell lines, the Molecular Libraries Small Molecule Repository (MLSMR) and cytotoxicity model identified ~10% of the hits in the top 1% screened (>10 fold enrichment). We also showed that seven out of nine Mtb active compounds from different academic published studies and eight out of eleven Mtb active compounds from a pharmaceutical screen (GSK) would have been identified by these Bayesian models.

CONCLUSION

Combining clustering and Bayesian models represents a useful strategy for compound prioritization and hit-to lead optimization of antitubercular agents.

Bayesian models leveraging bioactivity and cytotoxicity information for drug discovery

Identification of unique leads represents a significant challenge in drug discovery. This hurdle is magnified in neglected diseases such as tuberculosis. We have leveraged public high-throughput screening (HTS) data to experimentally validate a virtual screening approach employing Bayesian models built with bioactivity information (single-event model) as well as bioactivity and cytotoxicity information (dual-event model). We virtually screened a commercial library and experimentally confirmed actives with hit rates exceeding typical HTS results by one to two orders of magnitude. This initial dual-event Bayesian model identified compounds with antitubercular whole-cell activity and low mammalian cell cytotoxicity from a published set of antimalarials. The most potent hit exhibits the in vitro activity and in vitro/in vivo safety profile of a drug lead. These Bayesian models offer significant economies in time and cost to drug discovery.

Structure-activity relationships of 2-aminothiazoles effective against Mycobacterium tuberculosis

A series of 2-aminothiazoles was synthesized based on a HTS scaffold from a whole-cell screen against Mycobacterium tuberculosis (Mtb). The SAR shows the central thiazole moiety and the 2-pyridyl moiety at C-4 of the thiazole are intolerant to modification. However, the N-2 position of the aminothiazole exhibits high flexibility and we successfully improved the antitubercular activity of the initial hit by more than 128-fold through introduction of substituted benzoyl groups at this position. N-(3-Chlorobenzoyl)-4-(2-pyridinyl)-1,3-thiazol-2-amine (55) emerged as one of the most promising analogues with a MIC of 0.024μM or 0.008μg/mL in 7H9 media and therapeutic index of nearly ∼300. However, 55 is rapidly metabolized by human liver microsomes (t1/2=28min) with metabolism occurring at the invariant aminothiazole moiety and Mtb develops spontaneous low-level resistance with a frequency of ∼10(-5).

A medicinal chemists' guide to the unique difficulties of lead optimization for tuberculosis

Tuberculosis is a bacterial disease that predominantly affects the lungs and results in extensive tissue pathology. This pathology contributes to the complexity of drug development as it presents discrete microenvironments within which the bacterium resides, often under conditions where replication is limited and intrinsic drug susceptibility is low. This consolidated pathology also results in impaired vascularization that limits access of potential lead molecules to the site of infection. Translating these considerations into a target-product profile to guide lead optimization programs involves implementing unique in vitro and in vivo assays to maximize the likelihood of developing clinically meaningful candidates.

Computational models for neglected diseases: gaps and opportunities

Neglected diseases, such as Chagas disease, African sleeping sickness, and intestinal worms, affect millions of the world's poor. They disproportionately affect marginalized populations, lack effective treatments or vaccines, or existing products are not accessible to the populations affected. Computational approaches have been used across many of these diseases for various aspects of research or development, and yet data produced by computational approaches are not integrated and widely accessible to others. Here, we identify gaps in which computational approaches have been used for some neglected diseases and not others. We also make recommendations for the broad-spectrum integration of these techniques into a neglected disease drug discovery and development workflow.

Small molecule inhibitors of trans-translation have broad-spectrum antibiotic activity

The trans-translation pathway for protein tagging and ribosome release plays a critical role for viability and virulence in a wide range of pathogens but is not found in animals. To explore the use of trans-translation as a target for antibiotic development, a high-throughput screen and secondary screening assays were used to identify small molecule inhibitors of the pathway. Compounds that inhibited protein tagging and proteolysis of tagged proteins were recovered from the screen. One of the most active compounds, KKL-35, inhibited the trans-translation tagging reaction with an IC50 = 0.9 µM. KKL-35 and other compounds identified in the screen exhibited broad-spectrum antibiotic activity, validating trans-translation as a target for drug development. This unique target could play a key role in combating strains of pathogenic bacteria that are resistant to existing antibiotics.

Enhancing hit identification in Mycobacterium tuberculosis drug discovery using validated dual-event Bayesian models

High-throughput screening (HTS) in whole cells is widely pursued to find compounds active against Mycobacterium tuberculosis (Mtb) for further development towards new tuberculosis (TB) drugs. Hit rates from these screens, usually conducted at 10 to 25 µM concentrations, typically range from less than 1% to the low single digits. New approaches to increase the efficiency of hit identification are urgently needed to learn from past screening data. The pharmaceutical industry has for many years taken advantage of computational approaches to optimize compound libraries for in vitro testing, a practice not fully embraced by academic laboratories in the search for new TB drugs. Adapting these proven approaches, we have recently built and validated Bayesian machine learning models for predicting compounds with activity against Mtb based on publicly available large-scale HTS data from the Tuberculosis Antimicrobial Acquisition Coordinating Facility. We now demonstrate the largest prospective validation to date in which we computationally screened 82,403 molecules with these Bayesian models, assayed a total of 550 molecules in vitro, and identified 124 actives against Mtb. Individual hit rates for the different datasets varied from 15–28%. We have identified several FDA approved and late stage clinical candidate kinase inhibitors with activity against Mtb which may represent starting points for further optimization. The computational models developed herein and the commercially available molecules derived from them are now available to any group pursuing Mtb drug discovery.

A novel quinoline derivative that inhibits mycobacterial FtsZ

High throughput phenotypic screening of large commercially available libraries through two NIH programs has produced thousands of potentially interesting hits for further development as antitubercular agents. Unfortunately, these screens do not supply target information, and further follow up target identification is required to allow optimal rational design and development of highly active and selective clinical candidates. Cheminformatic analysis of the quinoline and quinazoline hits from these HTS screens suggested a hypothesis that certain compounds in these two classes may target the mycobacterial tubulin homolog, FtsZ. In this brief communication, activity of a lead quinoline against the target FtsZ from Mycobacterium tuberculosis (Mtb) is confirmed as well as good in vitro whole cell antibacterial activity against Mtb H37Rv. The identification of a putative target of this highly tractable pharmacophore should help medicinal chemists interested in targeting FtsZ and cell division develop a rational design program to optimize this activity toward a novel drug candidate.

A dual read-out assay to evaluate the potency of compounds active against Mycobacterium tuberculosis

Tuberculosis is a serious global health problem caused by the bacterium Mycobacterium tuberculosis. There is an urgent need for discovery and development of new treatments, but this can only be accomplished through rapid and reproducible M. tuberculosis assays designed to identify potent inhibitors. We developed an automated 96-well assay utilizing a recombinant strain of M. tuberculosis expressing a far-red fluorescent reporter to determine the activity of novel compounds; this allowed us to measure growth by monitoring both optical density and fluorescence. We determined that optical density and fluorescence were correlated with cell number during logarithmic phase growth. Fluorescence was stably maintained without antibiotic selection over 5 days, during which time cells remained actively growing. We optimized parameters for the assay, with the final format being 5 days' growth in 96-well plates in the presence of 2% w/v DMSO. We confirmed reproducibility using rifampicin and other antibiotics. The dual detection method allows for a reproducible calculation of the minimum inhibitory concentration (MIC), at the same time detecting artefacts such as fluorescence quenching or compound precipitation. We used our assay to confirm anti-tubercular activity and establish the structure activity relationship (SAR) around the imidazo[1,2-a]pyridine-3-carboxamides, a promising series of M. tuberculosis inhibitors.

6-Oxo and 6-thio purine analogs as antimycobacterial agents

6-Oxo and 6-thio analogs of purine were prepared based on the initial activity screening of a small, diverse purine library against Mycobacterium tuberculosis (Mtb). Certain 6-oxo and 6-thio-substituted purine analogs described herein showed moderate to good inhibitory activity. N(9)-substitution apparently enhances the anti-mycobacterial activity in the purine series described herein. Several 2-amino and 2-chloro purine analogs were also synthesized that showed moderate inhibitory activity against Mtb.

TB Mobile: a mobile app for anti-tuberculosis molecules with known targets

BACKGROUND

An increasing number of researchers are focused on strategies for developing inhibitors of Mycobacterium tuberculosis (Mtb) as tuberculosis (TB) drugs.

RESULTS

In order to learn from prior work we have collated information on molecules screened versus Mtb and their targets which has been made available in the Collaborative Drug Discovery (CDD) database. This dataset contains published data on target, essentiality, links to PubMed, TBDB, TBCyc (which provides a pathway-based visualization of the entire cellular biochemical network) and human homolog information. The development of mobile cheminformatics apps could lower the barrier to drug discovery and promote collaboration. Therefore we have used this set of over 700 molecules screened versus Mtb and their targets to create a free mobile app (TB Mobile) that displays molecule structures and links to the bioinformatics data. By input of a molecular structures and performing a similarity search within the app we can infer potential targets or search by targets to retrieve compounds known to be active.

CONCLUSIONS

TB Mobile may assist researchers as part of their workflow in identifying potential targets for hits generated from phenotypic screening and in prioritizing them for further follow-up. The app is designed to lower the barriers to accessing this information, so that all researchers with an interest in combatting this deadly disease can use it freely to the benefit of their own efforts.

A chemical proteomics approach to profiling the ATP-binding proteome of Mycobacterium tuberculosis

Tuberculosis, caused by Mycobacterium tuberculosis, remains one of the leading causes of death worldwide despite extensive research, directly observed therapy using multidrug regimens, and the widespread use of a vaccine. The majority of patients harbor the bacterium in a state of metabolic dormancy. New drugs with novel modes of action are needed to target essential metabolic pathways in M. tuberculosis; ATP-competitive enzyme inhibitors are one such class. Previous screening efforts for ATP-competitive enzyme inhibitors identified several classes of lead compounds that demonstrated potent anti-mycobacterial efficacy as well as tolerable levels of toxicity in cell culture. In this report, a probe-based chemoproteomic approach was used to selectively profile the M. tuberculosis ATP-binding proteome in normally growing and hypoxic M. tuberculosis. From these studies, 122 ATP-binding proteins were identified in either metabolic state, and roughly 60% of these are reported to be essential for survival in vitro. These data are available through ProteomeXchange with identifier PXD000141. Protein families vital to the survival of the tubercle bacillus during hypoxia emerged from our studies. Specifically, along with members of the DosR regulon, several proteins involved in energy metabolism (Icl/Rv0468 and Mdh/Rv1240) and lipid biosynthesis (UmaA/Rv0469, DesA1/Rv0824c, and DesA2/Rv1094) were found to be differentially abundant in hypoxic versus normal growing cultures. These pathways represent a subset of proteins that may be relevant therapeutic targets for development of novel ATP-competitive antibiotics.

Computational models for tuberculosis drug discovery

The search for small molecules with activity against Mycobacterium tuberculosis increasingly uses -high-throughput screening and computational methods. Previously, we have analyzed recent studies in which computational tools were used for cheminformatics. We have now updated this analysis to illustrate how they may assist in finding desirable leads for tuberculosis drug discovery. We provide our thoughts on strategies for drug discovery efforts for neglected diseases.

Identification of novel imidazo[1,2-a]pyridine inhibitors targeting M. tuberculosis QcrB

Mycobacterium tuberculosis is a major human pathogen and the causative agent for the pulmonary disease, tuberculosis (TB). Current treatment programs to combat TB are under threat due to the emergence of multi-drug and extensively-drug resistant TB. Through the use of high throughput whole cell screening of an extensive compound library a number of imidazo[1,2-a]pyridine (IP) compounds were obtained as potent lead molecules active against M. tuberculosis and Mycobacterium bovis BCG. The IP inhibitors (1-4) demonstrated minimum inhibitory concentrations (MICs) in the range of 0.03 to 5 µM against a panel of M. tuberculosis strains. M. bovis BCG spontaneous resistant mutants were generated against IP 1, 3, and 4 at 5× MIC and subsequent whole genome sequencing identified a single nucleotide polymorphism (937)ACC>(937)GCC (T313A) in qcrB, which encodes the b subunit of the electron transport ubiquinol cytochrome C reductase. This mutation also conferred cross-resistance against IP 1, 3 and 4 demonstrating a common target. Gene dosage experiments confirmed M. bovis BCG QcrB as the target where over-expression in M. bovis BCG led to an increase in MIC from 0.5 to >8 µM for IP 3. An acute murine model of TB infection established bacteriostatic activity of the IP series, which await further detailed characterization.

One pot solvent free synthesis and in vitro antitubercular screening of 3-Aracylphthalides against Mycobacterium tuberculosis

One pot synthesis of 3-Aracylphthalide was accomplished in good yield by reacting 2-carboxy benzaldehyde with various aromatic methyl ketones in presence of methane sulphonic acid. Various phthalides thus obtained were characterized with spectral techniques. These phthalides were subjected to in vitro antitubercular screening against Mycobacterium tuberculosis H37Ra (MTB) by using XRMA protocol. Among the phthalides screened, four exhibited half maximal inhibitory concentration (IC(50)) in the range of 0.81-1.24 μg/ml thereby providing potential lead compounds for future drug discovery studies.

Identification of novel Mt-Guab2 inhibitor series active against M. tuberculosis

Tuberculosis (TB) remains a leading cause of mortality worldwide. With the emergence of multidrug resistant TB, extensively drug resistant TB and HIV-associated TB it is imperative that new drug targets be identified. The potential of Mycobacterium tuberculosis inosine monophosphate dehydrogenase (IMPDH) as a novel drug target was explored in the present study. IMPDH exclusively catalyzes the conversion of inosine monophosphate (IMP) to xanthosine monophosphate (XMP) in the presence of the cofactor nicotinamide adenine dinucleotide (NAD⁺). Although the enzyme is a dehydrogenase, the enzyme does not catalyze the reverse reaction i.e. the conversion of XMP to IMP. Unlike other bacteria, M. tuberculosis harbors three IMPDH-like genes, designated as Mt-guaB1, Mt-guaB2 and Mt-guaB3 respectively. Of the three putative IMPDH's, we previously confirmed that Mt-GuaB2 was the only functional ortholog by characterizing the enzyme kinetically. Using an in silico approach based on designed scaffolds, a series of novel classes of inhibitors was identified. The inhibitors possess good activity against M. tuberculosis with MIC values in the range of 0.4 to 11.4 µg mL⁻¹. Among the identified ligands, two inhibitors have nanomolar K_is against the Mt-GuaB2 enzyme.