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

The Goldilocks paradigm: comparing classical machine learning, large language models, and few-shot learning for drug discovery applications

Recent advances in machine learning (ML) have led to newer model architectures including transformers (large language models, LLMs) showing state of the art results in text generation and image analysis as well as few-shot learning (FSLC) models which offer predictive power with extremely small datasets. These new architectures may offer promise, yet the 'no-free lunch' theorem suggests that no single model algorithm can outperform at all possible tasks. Here, we explore the capabilities of classical (SVR), FSLC, and transformer models (MolBART) over a range of dataset tasks and show a 'goldilocks zone' for each model type, in which dataset size and feature distribution (i.e. dataset "diversity") determines the optimal algorithm strategy. When datasets are small ( < 50 molecules), FSLC tend to outperform both classical ML and transformers. When datasets are small-to-medium sized (50-240 molecules) and diverse, transformers outperform both classical models and few-shot learning. Finally, when datasets are of larger and of sufficient size, classical models then perform the best, suggesting that the optimal model to choose likely depends on the dataset available, its size and diversity. These findings may help to answer the perennial question of which ML algorithm is to be used when faced with a new dataset.

Machine learning in medication prescription: A systematic review

BACKGROUND

Medication prescription is a complex process that could benefit from current research and development in machine learning through decision support systems. Particularly pediatricians are forced to prescribe medications "off-label" as children are still underrepresented in clinical studies, which leads to a high risk of an incorrect dose and adverse drug effects.

METHODS

PubMed, IEEE Xplore and PROSPERO were searched for relevant studies that developed and evaluated well-performing machine learning algorithms following the PRISMA statement. Quality assessment was conducted in accordance with the IJMEDI checklist. Identified studies were reviewed in detail, including the required variables for predicting the correct dose, especially of pediatric medication prescription.

RESULTS

The search identified 656 studies, of which 64 were reviewed in detail and 36 met the inclusion criteria. According to the IJMEDI checklist, five studies were considered to be of high quality. 19 of the 36 studies dealt with the active substance warfarin. Overall, machine learning algorithms based on decision trees or regression methods performed superior regarding their predictive power than algorithms based on neural networks, support vector machines or other methods. The use of ensemble methods like bagging or boosting generally enhanced the accuracy of the dose predictions. The required input and output variables of the algorithms were considerably heterogeneous and differ strongly among the respective substance.

CONCLUSIONS

By using machine learning algorithms, the prescription process could be simplified and dosing correctness could be enhanced. Despite the heterogenous results among the different substances and cases and the lack of pediatric use cases, the identified approaches and required variables can serve as an excellent starting point for further development of algorithms predicting drug doses, particularly for children. Especially the combination of physiologically-based pharmacokinetic models with machine learning algorithms represents a great opportunity to enhance the predictive power and accuracy of the developed algorithms.

High-Throughput Phenotypic Screening and Machine Learning Methods Enabled the Selection of Broad-Spectrum Low-Toxicity Antitrypanosomatidic Agents

Broad-spectrum anti-infective chemotherapy agents with activity against Trypanosomes, Leishmania, and Mycobacterium tuberculosis species were identified from a high-throughput phenotypic screening program of the 456 compounds belonging to the Ty-Box, an in-house industry database. Compound characterization using machine learning approaches enabled the identification and synthesis of 44 compounds with broad-spectrum antiparasitic activity and minimal toxicity against Trypanosoma brucei, Leishmania Infantum, and Trypanosoma cruzi. In vitro studies confirmed the predictive models identified in compound 40 which emerged as a new lead, featured by an innovative N-(5-pyrimidinyl)benzenesulfonamide scaffold and promising low micromolar activity against two parasites and low toxicity. Given the volume and complexity of data generated by the diverse high-throughput screening assays performed on the compounds of the Ty-Box library, the chemoinformatic and machine learning tools enabled the selection of compounds eligible for further evaluation of their biological and toxicological activities and aided in the decision-making process toward the design and optimization of the identified lead.

Machine Learning Models for Mycobacterium tuberculosisIn Vitro Activity: Prediction and Target Visualization

Tuberculosis (TB) is a major global health challenge, with approximately 1.4 million deaths per year. There is still a need to develop novel treatments for patients infected with Mycobacterium tuberculosis (Mtb). There have been many large-scale phenotypic screens that have led to the identification of thousands of new compounds. Yet, there is very limited investment in TB drug discovery which points to the need for new methods to increase the efficiency of drug discovery against Mtb. We have used machine learning approaches to learn from the public Mtb data, resulting in many data sets and models with robust enrichment and hit rates leading to the discovery of new active compounds. Recently, we have curated predominantly small-molecule Mtb data and developed new machine learning classification models with 18 886 molecules at different activity cutoffs. We now describe the further validation of these Bayesian models using a library of over 1000 molecules synthesized as part of EU-funded New Medicines for TB and More Medicines for TB programs. We highlight molecular features which are enriched in these active compounds. In addition, we provide new regression and classification models that can be used for scoring compound libraries or used to design new molecules. We have also visualized these molecules in the context of known molecular targets and identified clusters in chemical property space, which may aid in future target identification efforts. Finally, we are also making these data sets publicly available, representing a significant increase to the available Mtb inhibition data in the public domain.

Mycobacterium abscessus drug discovery using machine learning

The prevalence of infections by nontuberculous mycobacteria is increasing, having surpassed tuberculosis in the United States and much of the developed world. Nontuberculous mycobacteria occur naturally in the environment and are a significant problem for patients with underlying lung diseases such as bronchiectasis, chronic obstructive pulmonary disease, and cystic fibrosis. Current treatment regimens are lengthy, complicated, toxic and they are often unsuccessful as seen by disease recurrence. Mycobacterium abscessus is one of the most commonly encountered organisms in nontuberculous mycobacteria disease and it is the most difficult to eradicate. There is currently no systematically proven regimen that is effective for treating M. abscessus infections. Our approach to drug discovery integrates machine learning, medicinal chemistry and in vitro testing and has been previously applied to Mycobacterium tuberculosis. We have now identified several novel 1-(phenylsulfonyl)-1H-benzimidazol-2-amines that have weak activity on M. abscessus in vitro but may represent a starting point for future further medicinal chemistry optimization. We also address limitations still to be overcome with the machine learning approach for M. abscessus.

Quantum Machine Learning Algorithms for Drug Discovery Applications

The growing quantity of public and private data sets focused on small molecules screened against biological targets or whole organisms provides a wealth of drug discovery relevant data. This is matched by the availability of machine learning algorithms such as Support Vector Machines (SVM) and Deep Neural Networks (DNN) that are computationally expensive to perform on very large data sets with thousands of molecular descriptors. Quantum computer (QC) algorithms have been proposed to offer an approach to accelerate quantum machine learning over classical computer (CC) algorithms, however with significant limitations. In the case of cheminformatics, which is widely used in drug discovery, one of the challenges to overcome is the need for compression of large numbers of molecular descriptors for use on a QC. Here, we show how to achieve compression with data sets using hundreds of molecules (SARS-CoV-2) to hundreds of thousands of molecules (whole cell screening data sets for plague and M. tuberculosis) with SVM and the data reuploading classifier (a DNN equivalent algorithm) on a QC benchmarked against CC and hybrid approaches. This study illustrates the steps needed in order to be "quantum computer ready" in order to apply quantum computing to drug discovery and to provide the foundation on which to build this field.

Use of Artificial Intelligence and Machine Learning for Discovery of Drugs for Neglected Tropical Diseases

Neglected tropical diseases continue to create high levels of morbidity and mortality in a sizeable fraction of the world’s population, despite ongoing research into new treatments. Some of the most important technological developments that have accelerated drug discovery for diseases of affluent countries have not flowed down to neglected tropical disease drug discovery. Pharmaceutical development business models, cost of developing new drug treatments and subsequent costs to patients, and accessibility of technologies to scientists in most of the affected countries are some of the reasons for this low uptake and slow development relative to that for common diseases in developed countries. Computational methods are starting to make significant inroads into discovery of drugs for neglected tropical diseases due to the increasing availability of large databases that can be used to train ML models, increasing accuracy of these methods, lower entry barrier for researchers, and widespread availability of public domain machine learning codes. Here, the application of artificial intelligence, largely the subset called machine learning, to modelling and prediction of biological activities and discovery of new drugs for neglected tropical diseases is summarized. The pathways for the development of machine learning methods in the short to medium term and the use of other artificial intelligence methods for drug discovery is discussed. The current roadblocks to, and likely impacts of, synergistic new technological developments on the use of ML methods for neglected tropical disease drug discovery in the future are also discussed.

Bioactivity Comparison across Multiple Machine Learning Algorithms Using over 5000 Datasets for Drug Discovery

Machine learning methods are attracting considerable attention from the pharmaceutical industry for use in drug discovery and applications beyond. In recent studies, we and others have applied multiple machine learning algorithms and modeling metrics and, in some cases, compared molecular descriptors to build models for individual targets or properties on a relatively small scale. Several research groups have used large numbers of datasets from public databases such as ChEMBL in order to evaluate machine learning methods of interest to them. The largest of these types of studies used on the order of 1400 datasets. We have now extracted well over 5000 datasets from CHEMBL for use with the ECFP6 fingerprint and in comparison of our proprietary software Assay Central with random forest, k-nearest neighbors, support vector classification, naïve Bayesian, AdaBoosted decision trees, and deep neural networks (three layers). Model performance was assessed using an array of fivefold cross-validation metrics including area-under-the-curve, F1 score, Cohen's kappa, and Matthews correlation coefficient. Based on ranked normalized scores for the metrics or datasets, all methods appeared comparable, while the distance from the top indicated that Assay Central and support vector classification were comparable. Unlike prior studies which have placed considerable emphasis on deep neural networks (deep learning), no advantage was seen in this case. If anything, Assay Central may have been at a slight advantage as the activity cutoff for each of the over 5000 datasets representing over 570,000 unique compounds was based on Assay Central performance, although support vector classification seems to be a strong competitor. We also applied Assay Central to perform prospective predictions for the toxicity targets PXR and hERG to further validate these models. This work appears to be the largest scale comparison of these machine learning algorithms to date. Future studies will likely evaluate additional databases, descriptors, and machine learning algorithms and further refine the methods for evaluating and comparing such models.

Molecule Property Analyses of Active Compounds for Mycobacterium tuberculosis

Tuberculosis (TB) continues to claim the lives of around 1.7 million people per year. Most concerning are the reports of multidrug drug resistance. Paradoxically, this global health pandemic is demanding new therapies when resources and interest are waning. However, continued tuberculosis drug discovery is critical to address the global health need and burgeoning multidrug resistance. Many diverse classes of antitubercular compounds have been identified with activity in vitro and in vivo. Our analyses of over 100 active leads are representative of thousands of active compounds generated over the past decade, suggests that they come from few chemical classes or natural product sources. We are therefore repeatedly identifying compounds that are similar to those that preceded them. Our molecule-centered cheminformatics analyses point to the need to dramatically increase the diversity of chemical libraries tested and get outside of the historic Mtb property space if we are to generate novel improved antitubercular leads.

Cinnamic Derivatives as Antitubercular Agents: Characterization by Quantitative Structure-Activity Relationship Studies

Tuberculosis, caused by Mycobacterium tuberculosis (Mtb), remains one of the top ten causes of death worldwide and the main cause of mortality from a single infectious agent. The upsurge of multi- and extensively-drug resistant tuberculosis cases calls for an urgent need to develop new and more effective antitubercular drugs. As the cinnamoyl scaffold is a privileged and important pharmacophore in medicinal chemistry, some studies were conducted to find novel cinnamic acid derivatives (CAD) potentially active against tuberculosis. In this context, we have engaged in the setting up of a quantitative structure–activity relationships (QSAR) strategy to: (i) derive through multiple linear regression analysis a statistically significant model to describe the antitubercular activity of CAD towards wild-type Mtb; and (ii) identify the most relevant properties with an impact on the antitubercular behavior of those derivatives. The best-found model involved only geometrical and electronic CAD related properties and was successfully challenged through strict internal and external validation procedures. The physicochemical information encoded by the identified descriptors can be used to propose specific structural modifications to design better CAD antitubercular compounds.

Intrabacterial Metabolism Obscures the Successful Prediction of an InhA Inhibitor of Mycobacterium tuberculosis

Tuberculosis, caused by Mycobacterium tuberculosis (M. tuberculosis), kills 1.6 million people annually. To bridge the gap between structure- and cell-based drug discovery strategies, we are pioneering a computer-aided discovery paradigm that merges structure-based virtual screening with ligand-based, machine learning methods trained with cell-based data. This approach successfully identified N-(3-methoxyphenyl)-7-nitrobenzo[c][1,2,5]oxadiazol-4-amine (JSF-2164) as an inhibitor of purified InhA with whole-cell efficacy versus in vitro cultured M. tuberculosis. When the intrabacterial drug metabolism (IBDM) platform was leveraged, mechanistic studies demonstrated that JSF-2164 underwent a rapid F₄₂₀H₂-dependent biotransformation within M. tuberculosis to afford intrabacterial nitric oxide and two amines, identified as JSF-3616 and JSF-3617. Thus, metabolism of JSF-2164 obscured the InhA inhibition phenotype within cultured M. tuberculosis. This study demonstrates a new docking/Bayesian computational strategy to combine cell- and target-based drug screening and the need to probe intrabacterial metabolism when clarifying the antitubercular mechanism of action.

Concepts of Artificial Intelligence for Computer-Assisted Drug Discovery

Artificial intelligence (AI), and, in particular, deep learning as a subcategory of AI, provides opportunities for the discovery and development of innovative drugs. Various machine learning approaches have recently (re)emerged, some of which may be considered instances of domain-specific AI which have been successfully employed for drug discovery and design. This review provides a comprehensive portrayal of these machine learning techniques and of their applications in medicinal chemistry. After introducing the basic principles, alongside some application notes, of the various machine learning algorithms, the current state-of-the art of AI-assisted pharmaceutical discovery is discussed, including applications in structure- and ligand-based virtual screening, de novo drug design, physicochemical and pharmacokinetic property prediction, drug repurposing, and related aspects. Finally, several challenges and limitations of the current methods are summarized, with a view to potential future directions for AI-assisted drug discovery and design.

Survey of Machine Learning Techniques in Drug Discovery

Background:

Drug discovery, which is the process of discovering new candidate medications, is very important for pharmaceutical industries. At its current stage, discovering new drugs is still a very expensive and time-consuming process, requiring Phases I, II and III for clinical trials. Recently, machine learning techniques in Artificial Intelligence (AI), especially the deep learning techniques which allow a computational model to generate multiple layers, have been widely applied and achieved state-of-the-art performance in different fields, such as speech recognition, image classification, bioinformatics, etc. One very important application of these AI techniques is in the field of drug discovery.

Methods:

We did a large-scale literature search on existing scientific websites (e.g, ScienceDirect, Arxiv) and startup companies to understand current status of machine learning techniques in drug discovery.

Results:

Our experiments demonstrated that there are different patterns in machine learning fields and drug discovery fields. For example, keywords like prediction, brain, discovery, and treatment are usually in drug discovery fields. Also, the total number of papers published in drug discovery fields with machine learning techniques is increasing every year.

Conclusion:

The main focus of this survey is to understand the current status of machine learning techniques in the drug discovery field within both academic and industrial settings, and discuss its potential future applications. Several interesting patterns for machine learning techniques in drug discovery fields are discussed in this survey.

Comparing and Validating Machine Learning Models for Mycobacterium tuberculosis Drug Discovery

Tuberculosis is a global health dilemma. In 2016, the WHO reported 10.4 million incidences and 1.7 million deaths. The need to develop new treatments for those infected with Mycobacterium tuberculosis ( Mtb) has led to many large-scale phenotypic screens and many thousands of new active compounds identified in vitro. However, with limited funding, efforts to discover new active molecules against Mtb needs to be more efficient. Several computational machine learning approaches have been shown to have good enrichment and hit rates. We have curated small molecule Mtb data and developed new models with a total of 18,886 molecules with activity cutoffs of 10 μM, 1 μM, and 100 nM. These data sets were used to evaluate different machine learning methods (including deep learning) and metrics and to generate predictions for additional molecules published in 2017. One Mtb model, a combined in vitro and in vivo data Bayesian model at a 100 nM activity yielded the following metrics for 5-fold cross validation: accuracy = 0.88, precision = 0.22, recall = 0.91, specificity = 0.88, kappa = 0.31, and MCC = 0.41. We have also curated an evaluation set ( n = 153 compounds) published in 2017, and when used to test our model, it showed the comparable statistics (accuracy = 0.83, precision = 0.27, recall = 1.00, specificity = 0.81, kappa = 0.36, and MCC = 0.47). We have also compared these models with additional machine learning algorithms showing Bayesian machine learning models constructed with literature Mtb data generated by different laboratories generally were equivalent to or outperformed deep neural networks with external test sets. Finally, we have also compared our training and test sets to show they were suitably diverse and different in order to represent useful evaluation sets. Such Mtb machine learning models could help prioritize compounds for testing in vitro and in vivo.

Comparison of Deep Learning With Multiple Machine Learning Methods and Metrics Using Diverse Drug Discovery Data Sets

Machine learning methods have been applied to many data sets in pharmaceutical research for several decades. The relative ease and availability of fingerprint type molecular descriptors paired with Bayesian methods resulted in the widespread use of this approach for a diverse array of end points relevant to drug discovery. Deep learning is the latest machine learning algorithm attracting attention for many of pharmaceutical applications from docking to virtual screening. Deep learning is based on an artificial neural network with multiple hidden layers and has found considerable traction for many artificial intelligence applications. We have previously suggested the need for a comparison of different machine learning methods with deep learning across an array of varying data sets that is applicable to pharmaceutical research. End points relevant to pharmaceutical research include absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) properties, as well as activity against pathogens and drug discovery data sets. In this study, we have used data sets for solubility, probe-likeness, hERG, KCNQ1, bubonic plague, Chagas, tuberculosis, and malaria to compare different machine learning methods using FCFP6 fingerprints. These data sets represent whole cell screens, individual proteins, physicochemical properties as well as a data set with a complex end point. Our aim was to assess whether deep learning offered any improvement in testing when assessed using an array of metrics including AUC, F1 score, Cohen's kappa, Matthews correlation coefficient and others. Based on ranked normalized scores for the metrics or data sets Deep Neural Networks (DNN) ranked higher than SVM, which in turn was ranked higher than all the other machine learning methods. Visualizing these properties for training and test sets using radar type plots indicates when models are inferior or perhaps over trained. These results also suggest the need for assessing deep learning further using multiple metrics with much larger scale comparisons, prospective testing as well as assessment of different fingerprints and DNN architectures beyond those used.

Collaborative drug discovery for More Medicines for Tuberculosis (MM4TB)

Neglected disease drug discovery is generally poorly funded compared with major diseases and hence there is an increasing focus on collaboration and precompetitive efforts such as public-private partnerships (PPPs). The More Medicines for Tuberculosis (MM4TB) project is one such collaboration funded by the EU with the goal of discovering new drugs for tuberculosis. Collaborative Drug Discovery has provided a commercial web-based platform called CDD Vault which is a hosted collaborative solution for securely sharing diverse chemistry and biology data. Using CDD Vault alongside other commercial and free cheminformatics tools has enabled support of this and other large collaborative projects, aiding drug discovery efforts and fostering collaboration. We will describe CDD's efforts in assisting with the MM4TB project.

The Next Era: Deep Learning in Pharmaceutical Research

Over the past decade we have witnessed the increasing sophistication of machine learning algorithms applied in daily use from internet searches, voice recognition, social network software to machine vision software in cameras, phones, robots and self-driving cars. Pharmaceutical research has also seen its fair share of machine learning developments. For example, applying such methods to mine the growing datasets that are created in drug discovery not only enables us to learn from the past but to predict a molecule's properties and behavior in future. The latest machine learning algorithm garnering significant attention is deep learning, which is an artificial neural network with multiple hidden layers. Publications over the last 3 years suggest that this algorithm may have advantages over previous machine learning methods and offer a slight but discernable edge in predictive performance. The time has come for a balanced review of this technique but also to apply machine learning methods such as deep learning across a wider array of endpoints relevant to pharmaceutical research for which the datasets are growing such as physicochemical property prediction, formulation prediction, absorption, distribution, metabolism, excretion and toxicity (ADME/Tox), target prediction and skin permeation, etc. We also show that there are many potential applications of deep learning beyond cheminformatics. It will be important to perform prospective testing (which has been carried out rarely to date) in order to convince skeptics that there will be benefits from investing in this technique.

Machine Learning Model Analysis and Data Visualization with Small Molecules Tested in a Mouse Model of Mycobacterium tuberculosis Infection (2014-2015)

The renewed urgency to develop new treatments for Mycobacterium tuberculosis (Mtb) infection has resulted in large-scale phenotypic screening and thousands of new active compounds in vitro. The next challenge is to identify candidates to pursue in a mouse in vivo efficacy model as a step to predicting clinical efficacy. We previously analyzed over 70 years of this mouse in vivo efficacy data, which we used to generate and validate machine learning models. Curation of 60 additional small molecules with in vivo data published in 2014 and 2015 was undertaken to further test these models. This represents a much larger test set than for the previous models. Several computational approaches have now been applied to analyze these molecules and compare their molecular properties beyond those attempted previously. Our previous machine learning models have been updated, and a novel aspect has been added in the form of mouse liver microsomal half-life (MLM t_1/2) and in vitro-based Mtb models incorporating cytotoxicity data that were used to predict in vivo activity for comparison. Our best Mtbin vivo models possess fivefold ROC values > 0.7, sensitivity > 80%, and concordance > 60%, while the best specificity value is >40%. Use of an MLM t_1/2 Bayesian model affords comparable results for scoring the 60 compounds tested. Combining MLM stability and in vitroMtb models in a novel consensus workflow in the best cases has a positive predicted value (hit rate) > 77%. Our results indicate that Bayesian models constructed with literature in vivoMtb data generated by different laboratories in various mouse models can have predictive value and may be used alongside MLM t_1/2 and in vitro-based Mtb models to assist in selecting antitubercular compounds with desirable in vivo efficacy. We demonstrate for the first time that consensus models of any kind can be used to predict in vivo activity for Mtb. In addition, we describe a new clustering method for data visualization and apply this to the in vivo training and test data, ultimately making the method accessible in a mobile app.

Predicting Mouse Liver Microsomal Stability with "Pruned" Machine Learning Models and Public Data

PURPOSE

Mouse efficacy studies are a critical hurdle to advance translational research of potential therapeutic compounds for many diseases. Although mouse liver microsomal (MLM) stability studies are not a perfect surrogate for in vivo studies of metabolic clearance, they are the initial model system used to assess metabolic stability. Consequently, we explored the development of machine learning models that can enhance the probability of identifying compounds possessing MLM stability.

METHODS

Published assays on MLM half-life values were identified in PubChem, reformatted, and curated to create a training set with 894 unique small molecules. These data were used to construct machine learning models assessed with internal cross-validation, external tests with a published set of antitubercular compounds, and independent validation with an additional diverse set of 571 compounds (PubChem data on percent metabolism).

RESULTS

"Pruning" out the moderately unstable / moderately stable compounds from the training set produced models with superior predictive power. Bayesian models displayed the best predictive power for identifying compounds with a half-life ≥1 h.

CONCLUSIONS

Our results suggest the pruning strategy may be of general benefit to improve test set enrichment and provide machine learning models with enhanced predictive value for the MLM stability of small organic molecules. This study represents the most exhaustive study to date of using machine learning approaches with MLM data from public sources.

Open Source Bayesian Models. 3. Composite Models for Prediction of Binned Responses

Bayesian models constructed from structure-derived fingerprints have been a popular and useful method for drug discovery research when applied to bioactivity measurements that can be effectively classified as active or inactive. The results can be used to rank candidate structures according to their probability of activity, and this ranking benefits from the high degree of interpretability when structure-based fingerprints are used, making the results chemically intuitive. Besides selecting an activity threshold, building a Bayesian model is fast and requires few or no parameters or user intervention. The method also does not suffer from such acute overtraining problems as quantitative structure–activity relationships or quantitative structure–property relationships (QSAR/QSPR). This makes it an approach highly suitable for automated workflows that are independent of user expertise or prior knowledge of the training data. We now describe a new method for creating a composite group of Bayesian models to extend the method to work with multiple states, rather than just binary. Incoming activities are divided into bins, each covering a mutually exclusive range of activities. For each of these bins, a Bayesian model is created to model whether or not the compound belongs in the bin. Analyzing putative molecules using the composite model involves making a prediction for each bin and examining the relative likelihood for each assignment, for example, highest value wins. The method has been evaluated on a collection of hundreds of data sets extracted from ChEMBL v20 and validated data sets for ADME/Tox and bioactivity.

Machine learning models identify molecules active against the Ebola virus in vitro

The search for small molecule inhibitors of Ebola virus (EBOV) has led to several high throughput screens over the past 3 years. These have identified a range of FDA-approved active pharmaceutical ingredients (APIs) with anti-EBOV activity in vitro and several of which are also active in a mouse infection model. There are millions of additional commercially-available molecules that could be screened for potential activities as anti-EBOV compounds. One way to prioritize compounds for testing is to generate computational models based on the high throughput screening data and then virtually screen compound libraries. In the current study, we have generated Bayesian machine learning models with viral pseudotype entry assay and the EBOV replication assay data. We have validated the models internally and externally. We have also used these models to computationally score the MicroSource library of drugs to select those likely to be potential inhibitors. Three of the highest scoring molecules that were not in the model training sets, quinacrine, pyronaridine and tilorone, were tested in vitro and had EC50 values of 350, 420 and 230 nM, respectively. Pyronaridine is a component of a combination therapy for malaria that was recently approved by the European Medicines Agency, which may make it more readily accessible for clinical testing. Like other known antimalarial drugs active against EBOV, it shares the 4-aminoquinoline scaffold. Tilorone, is an investigational antiviral agent that has shown a broad array of biological activities including cell growth inhibition in cancer cells, antifibrotic properties, α7 nicotinic receptor agonist activity, radioprotective activity and activation of hypoxia inducible factor-1. Quinacrine is an antimalarial but also has use as an anthelmintic. Our results suggest data sets with less than 1,000 molecules can produce validated machine learning models that can in turn be utilized to identify novel EBOV inhibitors in vitro.

Combining Metabolite-Based Pharmacophores with Bayesian Machine Learning Models for Mycobacterium tuberculosis Drug Discovery

Integrated computational approaches for Mycobacterium tuberculosis (Mtb) are useful to identify new molecules that could lead to future tuberculosis (TB) drugs. Our approach uses information derived from the TBCyc pathway and genome database, the Collaborative Drug Discovery TB database combined with 3D pharmacophores and dual event Bayesian models of whole-cell activity and lack of cytotoxicity. We have prioritized a large number of molecules that may act as mimics of substrates and metabolites in the TB metabolome. We computationally searched over 200,000 commercial molecules using 66 pharmacophores based on substrates and metabolites from Mtb and further filtering with Bayesian models. We ultimately tested 110 compounds in vitro that resulted in two compounds of interest, BAS 04912643 and BAS 00623753 (MIC of 2.5 and 5 μg/mL, respectively). These molecules were used as a starting point for hit-to-lead optimization. The most promising class proved to be the quinoxaline di-N-oxides, evidenced by transcriptional profiling to induce mRNA level perturbations most closely resembling known protonophores. One of these, SRI58 exhibited an MIC = 1.25 μg/mL versus Mtb and a CC₅₀ in Vero cells of >40 μg/mL, while featuring fair Caco-2 A-B permeability (2.3 x 10⁻⁶ cm/s), kinetic solubility (125 μM at pH 7.4 in PBS) and mouse metabolic stability (63.6% remaining after 1 h incubation with mouse liver microsomes). Despite demonstration of how a combined bioinformatics/cheminformatics approach afforded a small molecule with promising in vitro profiles, we found that SRI58 did not exhibit quantifiable blood levels in mice.

Machine learning models identify molecules active against the Ebola virus in vitro

The search for small molecule inhibitors of Ebola virus (EBOV) has led to several high throughput screens over the past 3 years. These have identified a range of FDA-approved active pharmaceutical ingredients (APIs) with anti-EBOV activity in vitro and several of which are also active in a mouse infection model. There are millions of additional commercially-available molecules that could be screened for potential activities as anti-EBOV compounds. One way to prioritize compounds for testing is to generate computational models based on the high throughput screening data and then virtually screen compound libraries. In the current study, we have generated Bayesian machine learning models with viral pseudotype entry assay and the EBOV replication assay data. We have validated the models internally and externally. We have also used these models to computationally score the MicroSource library of drugs to select those likely to be potential inhibitors. Three of the highest scoring molecules that were not in the model training sets, quinacrine, pyronaridine and tilorone, were tested in vitro and had EC ₅₀ values of 350, 420 and 230 nM, respectively. Pyronaridine is a component of a combination therapy for malaria that was recently approved by the European Medicines Agency, which may make it more readily accessible for clinical testing. Like other known antimalarial drugs active against EBOV, it shares the 4-aminoquinoline scaffold. Tilorone, is an investigational antiviral agent that has shown a broad array of biological activities including cell growth inhibition in cancer cells, antifibrotic properties, α7 nicotinic receptor agonist activity, radioprotective activity and activation of hypoxia inducible factor-1. Quinacrine is an antimalarial but also has use as an anthelmintic. Our results suggest data sets with less than 1,000 molecules can produce validated machine learning models that can in turn be utilized to identify novel EBOV inhibitors in vitro.

Machine learning models identify molecules active against the Ebola virus in vitro

The search for small molecule inhibitors of Ebola virus (EBOV) has led to several high throughput screens over the past 3 years. These have identified a range of FDA-approved active pharmaceutical ingredients (APIs) with anti-EBOV activity in vitro and several of which are also active in a mouse infection model. There are millions of additional commercially-available molecules that could be screened for potential activities as anti-EBOV compounds. One way to prioritize compounds for testing is to generate computational models based on the high throughput screening data and then virtually screen compound libraries. In the current study, we have generated Bayesian machine learning models with viral pseudotype entry assay and the EBOV replication assay data. We have validated the models internally and externally. We have also used these models to computationally score the MicroSource library of drugs to select those likely to be potential inhibitors. Three of the highest scoring molecules that were not in the model training sets, quinacrine, pyronaridine and tilorone, were tested in vitro and had EC ₅₀ values of 350, 420 and 230 nM, respectively. Pyronaridine is a component of a combination therapy for malaria that was recently approved by the European Medicines Agency, which may make it more readily accessible for clinical testing. Like other known antimalarial drugs active against EBOV, it shares the 4-aminoquinoline scaffold. Tilorone, is an investigational antiviral agent that has shown a broad array of biological activities including cell growth inhibition in cancer cells, antifibrotic properties, α7 nicotinic receptor agonist activity, radioprotective activity and activation of hypoxia inducible factor-1. Quinacrine is an antimalarial but also has use as an anthelmintic. Our results suggest data sets with less than 1,000 molecules can produce validated machine learning models that can in turn be utilized to identify novel EBOV inhibitors in vitro.

Machine Learning Models and Pathway Genome Data Base for Trypanosoma cruzi Drug Discovery

Background

Chagas disease is a neglected tropical disease (NTD) caused by the eukaryotic parasite Trypanosoma cruzi. The current clinical and preclinical pipeline for T. cruzi is extremely sparse and lacks drug target diversity.

Methodology/Principal Findings

In the present study we developed a computational approach that utilized data from several public whole-cell, phenotypic high throughput screens that have been completed for T. cruzi by the Broad Institute, including a single screen of over 300,000 molecules in the search for chemical probes as part of the NIH Molecular Libraries program. We have also compiled and curated relevant biological and chemical compound screening data including (i) compounds and biological activity data from the literature, (ii) high throughput screening datasets, and (iii) predicted metabolites of T. cruzi metabolic pathways. This information was used to help us identify compounds and their potential targets. We have constructed a Pathway Genome Data Base for T. cruzi. In addition, we have developed Bayesian machine learning models that were used to virtually screen libraries of compounds. Ninety-seven compounds were selected for in vitro testing, and 11 of these were found to have EC₅₀ < 10μM. We progressed five compounds to an in vivo mouse efficacy model of Chagas disease and validated that the machine learning model could identify in vitro active compounds not in the training set, as well as known positive controls. The antimalarial pyronaridine possessed 85.2% efficacy in the acute Chagas mouse model. We have also proposed potential targets (for future verification) for this compound based on structural similarity to known compounds with targets in T. cruzi.

Conclusions/ Significance

We have demonstrated how combining chemoinformatics and bioinformatics for T. cruzi drug discovery can bring interesting in vivo active molecules to light that may have been overlooked. The approach we have taken is broadly applicable to other NTDs.

Chagas disease is a neglected tropical disease (NTD) caused by the eukaryotic parasite Trypanosoma cruzi. The disease is endemic to Latin America but is increasingly found in North America and Europe, primarily through immigration, and the spread of this disease is bringing new attention to the need for novel, safe, and effective therapeutics to treat T. cruzi infection. We have used data from a phenotypic screen to build Bayesian models to predict anti-parasitic activity against T. cruzi in vitro. These models were used to score various small libraries of molecules. We selected less than 100 compounds for testing and found in vitro actives, some of which were tested in an in vivo efficacy model. We identified the antimalarial pyronaridine as having in vivo efficacy and provides us with a new starting point for further investigation and optimization.

Open Source Bayesian Models. 1. Application to ADME/Tox and Drug Discovery Datasets

On the order of hundreds of absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) models have been described in the literature in the past decade which are more often than not inaccessible to anyone but their authors. Public accessibility is also an issue with computational models for bioactivity, and the ability to share such models still remains a major challenge limiting drug discovery. We describe the creation of a reference implementation of a Bayesian model-building software module, which we have released as an open source component that is now included in the Chemistry Development Kit (CDK) project, as well as implemented in the CDD Vault and in several mobile apps. We use this implementation to build an array of Bayesian models for ADME/Tox, in vitro and in vivo bioactivity, and other physicochemical properties. We show that these models possess cross-validation receiver operator curve values comparable to those generated previously in prior publications using alternative tools. We have now described how the implementation of Bayesian models with FCFP6 descriptors generated in the CDD Vault enables the rapid production of robust machine learning models from public data or the user’s own datasets. The current study sets the stage for generating models in proprietary software (such as CDD) and exporting these models in a format that could be run in open source software using CDK components. This work also demonstrates that we can enable biocomputation across distributed private or public datasets to enhance drug discovery.

Open Source Bayesian Models. 2. Mining a "Big Dataset" To Create and Validate Models with ChEMBL

In an associated paper, we have described a reference implementation of Laplacian-corrected naïve Bayesian model building using extended connectivity (ECFP)- and molecular function class fingerprints of maximum diameter 6 (FCFP)-type fingerprints. As a follow-up, we have now undertaken a large-scale validation study in order to ensure that the technique generalizes to a broad variety of drug discovery datasets. To achieve this, we have used the ChEMBL (version 20) database and split it into more than 2000 separate datasets, each of which consists of compounds and measurements with the same target and activity measurement. In order to test these datasets with the two-state Bayesian classification, we developed an automated algorithm for detecting a suitable threshold for active/inactive designation, which we applied to all collections. With these datasets, we were able to establish that our Bayesian model implementation is effective for the large majority of cases, and we were able to quantify the impact of fingerprint folding on the receiver operator curve cross-validation metrics. We were also able to study the impact that the choice of training/testing set partitioning has on the resulting recall rates. The datasets have been made publicly available to be downloaded, along with the corresponding model data files, which can be used in conjunction with the CDK and several mobile apps. We have also explored some novel visualization methods which leverage the structural origins of the ECFP/FCFP fingerprints to attribute regions of a molecule responsible for positive and negative contributions to activity. The ability to score molecules across thousands of relevant datasets across organisms also may help to access desirable and undesirable off-target effects as well as suggest potential targets for compounds derived from phenotypic screens.

Finding small molecules for the 'next Ebola'

The current Ebola virus epidemic may provide some suggestions of how we can better prepare for the next pathogen outbreak. We propose several cost effective steps that could be taken that would impact the discovery and use of small molecule therapeutics including: 1. text mine the literature, 2. patent assignees and/or inventors should openly declare their relevant filings, 3. reagents and assays could be commoditized, 4. using manual curation to enhance database links, 5. engage database and curation teams, 6. consider open science approaches, 7. adapt the "box" model for shareable reference compounds, and 8. involve the physician's perspective.

Finding small molecules for the 'next Ebola'

The current Ebola virus epidemic may provide some suggestions of how we can better prepare for the next pathogen outbreak. We propose several cost effective steps that could be taken that would impact the discovery and use of small molecule therapeutics including: 1. text mine the literature, 2. patent assignees and/or inventors should openly declare their relevant filings, 3. reagents and assays could be commoditized, 4. using manual curation to enhance database links, 5. engage database and curation teams, 6. consider open science approaches, 7. adapt the "box" model for shareable reference compounds, and 8. involve the physician's perspective.

Bigger data, collaborative tools and the future of predictive drug discovery

Over the past decade we have seen a growth in the provision of chemistry data and cheminformatics tools as either free websites or software as a service commercial offerings. These have transformed how we find molecule-related data and use such tools in our research. There have also been efforts to improve collaboration between researchers either openly or through secure transactions using commercial tools. A major challenge in the future will be how such databases and software approaches handle larger amounts of data as it accumulates from high throughput screening and enables the user to draw insights, enable predictions and move projects forward. We now discuss how information from some drug discovery datasets can be made more accessible and how privacy of data should not overwhelm the desire to share it at an appropriate time with collaborators. We also discuss additional software tools that could be made available and provide our thoughts on the future of predictive drug discovery in this age of big data. We use some examples from our own research on neglected diseases, collaborations, mobile apps and algorithm development to illustrate these ideas.