Machine-learning approaches in drug discovery: methods and applications

An Analysis of QSAR Research Based on Machine Learning Concepts

Quantitative Structure-Activity Relationship (QSAR) is a popular approach developed to correlate chemical molecules with their biological activities based on their chemical structures. Machine learning techniques have proved to be promising solutions to QSAR modeling. Due to the significant role of machine learning strategies in QSAR modeling, this area of research has attracted much attention from researchers. A considerable amount of literature has been published on machine learning based QSAR modeling methodologies whilst this domain still suffers from lack of a recent and comprehensive analysis of these algorithms. This study systematically reviews the application of machine learning algorithms in QSAR, aiming to provide an analytical framework. For this purpose, we present a framework called 'ML-QSAR'. This framework has been designed for future research to: a) facilitate the selection of proper strategies among existing algorithms according to the application area requirements, b) help to develop and ameliorate current methods and c) providing a platform to study existing methodologies comparatively. In ML-QSAR, first a structured categorization is depicted which studied the QSAR modeling research based on machine models. Then several criteria are introduced in order to assess the models. Finally, inspired by aforementioned criteria the qualitative analysis is carried out.

Generative Deep Neural Networks for Inverse Materials Design Using Backpropagation and Active Learning

In recent years, machine learning (ML) techniques are seen to be promising tools to discover and design novel materials. However, the lack of robust inverse design approaches to identify promising candidate materials without exploring the entire design space causes a fundamental bottleneck. A general-purpose inverse design approach is presented using generative inverse design networks. This ML-based inverse design approach uses backpropagation to calculate the analytical gradients of an objective function with respect to design variables. This inverse design approach is capable of overcoming local minima traps by using backpropagation to provide rapid calculations of gradient information and running millions of optimizations with different initial values. Furthermore, an active learning strategy is adopted in the inverse design approach to improve the performance of candidate materials and reduce the amount of training data needed to do so. Compared to passive learning, the active learning strategy is capable of generating better designs and reducing the amount of training data by at least an order-of-magnitude in the case study on composite materials. The inverse design approach is compared with conventional gradient-based topology optimization and gradient-free genetic algorithms and the pros and cons of each method are discussed when applied to materials discovery and design problems.

Automated exploration of the low-energy chemical space with fast quantum chemical methods

We propose and discuss an efficient scheme for the in silico sampling for parts of the molecular chemical space by semiempirical tight-binding methods combined with a meta-dynamics driven search algorithm. The focus of this work is set on the generation of proper thermodynamic ensembles at a quantum chemical level for conformers, but similar procedures for protonation states, tautomerism and non-covalent complex geometries are also discussed. The conformational ensembles consisting of all significantly populated minimum energy structures normally form the basis of further, mostly DFT computational work, such as the calculation of spectra or macroscopic properties. By using basic quantum chemical methods, electronic effects or possible bond breaking/formation are accounted for and a very reasonable initial energetic ranking of the candidate structures is obtained. Due to the huge computational speedup gained by the fast low-cost quantum chemical methods, overall short computation times even for systems with hundreds of atoms (typically drug-sized molecules) are achieved. Furthermore, specialized applications, such as sampling with implicit solvation models or constrained conformational sampling for transition-states, metal-, surface-, or noncovalently bound complexes are discussed, opening many possible applications in modern computational chemistry and drug discovery. The procedures have been implemented in a freely available computer code called CREST, that makes use of the fast and reliable GFNn-xTB methods.

Applying Machine Learning to Ultrafast Shape Recognition in Ligand-Based Virtual Screening

Ultrafast Shape Recognition (USR), along with its derivatives, are Ligand-Based Virtual Screening (LBVS) methods that condense 3-dimensional information about molecular shape, as well as other properties, into a small set of numeric descriptors. These can be used to efficiently compute a measure of similarity between pairs of molecules using a simple inverse Manhattan Distance metric. In this study we explore the use of suitable Machine Learning techniques that can be trained using USR descriptors, so as to improve the similarity detection of potential new leads. We use molecules from the Directory for Useful Decoys-Enhanced to construct machine learning models based on three different algorithms: Gaussian Mixture Models (GMMs), Isolation Forests and Artificial Neural Networks (ANNs). We train models based on full molecule conformer models, as well as the Lowest Energy Conformations (LECs) only. We also investigate the performance of our models when trained on smaller datasets so as to model virtual screening scenarios when only a small number of actives are known a priori. Our results indicate significant performance gains over a state of the art USR-derived method, ElectroShape 5D, with GMMs obtaining a mean performance up to 430% better than that of ElectroShape 5D in terms of Enrichment Factor with a maximum improvement of up to 940%. Additionally, we demonstrate that our models are capable of maintaining their performance, in terms of enrichment factor, within 10% of the mean as the size of the training dataset is successively reduced. Furthermore, we also demonstrate that running times for retrospective screening using the machine learning models we selected are faster than standard USR, on average by a factor of 10, including the time required for training. Our results show that machine learning techniques can significantly improve the virtual screening performance and efficiency of the USR family of methods.

A Pilot Study of All-Computational Drug Design Protocol-From Structure Prediction to Interaction Analysis

Speeding up the drug discovery process is of great significance. To achieve that, high-efficiency methods should be exploited. The conventional wet-bench methods hardly meet the high-speed demand due to time-consuming experiments. Conversely, in silico approaches are much more efficient for drug discovery and design. However, in silico approaches usually serve as a supportive role in research processes. To fully exert the strength of computational methods, we propose a protocol which integrates various in silico approaches, from de novo protein structure prediction to ligand-protein interaction simulation. As a proof of concept, human SK2/calmodulin complex was used as a target for validation. First, we obtained a predicted structure of SK2/calmodulin and predicted binding sites which were consistent with the literature data. Then we investigated the ligand-protein interaction via virtual mutagenesis, flexible docking, and binding affinity calculation. As a result, the binding energies of mutants have similar trends compared with the EC₅₀ values (R = 0.6 for NS309 in V481 mutants). The results indicate that our protocol can be applied to the drug design of structure unknown proteins. Our study also demonstrates that the integration of in silico approaches is feasible and it facilitates the acceleration of new drug discovery.

The impact of chemoinformatics on drug discovery in the pharmaceutical industry

Introduction: Even though there have been substantial advances in our understanding of biological systems, research in drug discovery is only just now beginning to utilize this type of information. The single-target paradigm, which exemplifies the reductionist approach, remains a mainstay of drug research today. A deeper view of the complexity involved in drug discovery is necessary to advance on this field.Areas covered: This perspective provides a summary of research areas where cheminformatics has played a key role in drug discovery, including of the available resources as well as a personal perspective of the challenges still faced in the field.Expert opinion: Although great strides have been made in the handling and analysis of biological and pharmacological data, more must be done to link the data to biological pathways. This is crucial if one is to understand how drugs modify disease phenotypes, although this will involve a shift from the single drug/single target paradigm that remains a mainstay of drug research. Moreover, such a shift would require an increased awareness of the role of physiology in the mechanism of drug action, which will require the introduction of new mathematical, computer, and biological methods for chemoinformaticians to be trained in.

Machine learning approaches and databases for prediction of drug-target interaction: a survey paper

The task of predicting the interactions between drugs and targets plays a key role in the process of drug discovery. There is a need to develop novel and efficient prediction approaches in order to avoid costly and laborious yet not-always-deterministic experiments to determine drug–target interactions (DTIs) by experiments alone. These approaches should be capable of identifying the potential DTIs in a timely manner. In this article, we describe the data required for the task of DTI prediction followed by a comprehensive catalog consisting of machine learning methods and databases, which have been proposed and utilized to predict DTIs. The advantages and disadvantages of each set of methods are also briefly discussed. Lastly, the challenges one may face in prediction of DTI using machine learning approaches are highlighted and we conclude by shedding some lights on important future research directions.

Secure multiparty computation for privacy-preserving drug discovery

Motivation

Quantitative structure–activity relationship (QSAR) and drug–target interaction (DTI) prediction are both commonly used in drug discovery. Collaboration among pharmaceutical institutions can lead to better performance in both QSAR and DTI prediction. However, the drug-related data privacy and intellectual property issues have become a noticeable hindrance for inter-institutional collaboration in drug discovery.

Results

We have developed two novel algorithms under secure multiparty computation (MPC), including QSARMPC and DTIMPC, which enable pharmaceutical institutions to achieve high-quality collaboration to advance drug discovery without divulging private drug-related information. QSARMPC, a neural network model under MPC, displays good scalability and performance and is feasible for privacy-preserving collaboration on large-scale QSAR prediction. DTIMPC integrates drug-related heterogeneous network data and accurately predicts novel DTIs, while keeping the drug information confidential. Under several experimental settings that reflect the situations in real drug discovery scenarios, we have demonstrated that DTIMPC possesses significant performance improvement over the baseline methods, generates novel DTI predictions with supporting evidence from the literature and shows the feasible scalability to handle growing DTI data. All these results indicate that QSARMPC and DTIMPC can provide practically useful tools for advancing privacy-preserving drug discovery.

Availability and implementation

The source codes of QSARMPC and DTIMPC are available on the GitHub: https://github.com/rongma6/QSARMPC_DTIMPC.git.

Supplementary information

Supplementary data are available at Bioinformatics online.

Machine-Learning-Assisted De Novo Design of Organic Molecules and Polymers: Opportunities and Challenges

Organic molecules and polymers have a broad range of applications in biomedical, chemical, and materials science fields. Traditional design approaches for organic molecules and polymers are mainly experimentally-driven, guided by experience, intuition, and conceptual insights. Though they have been successfully applied to discover many important materials, these methods are facing significant challenges due to the tremendous demand of new materials and vast design space of organic molecules and polymers. Accelerated and inverse materials design is an ideal solution to these challenges. With advancements in high-throughput computation, artificial intelligence (especially machining learning, ML), and the growth of materials databases, ML-assisted materials design is emerging as a promising tool to flourish breakthroughs in many areas of materials science and engineering. To date, using ML-assisted approaches, the quantitative structure property/activity relation for material property prediction can be established more accurately and efficiently. In addition, materials design can be revolutionized and accelerated much faster than ever, through ML-enabled molecular generation and inverse molecular design. In this perspective, we review the recent progresses in ML-guided design of organic molecules and polymers, highlight several successful examples, and examine future opportunities in biomedical, chemical, and materials science fields. We further discuss the relevant challenges to solve in order to fully realize the potential of ML-assisted materials design for organic molecules and polymers. In particular, this study summarizes publicly available materials databases, feature representations for organic molecules, open-source tools for feature generation, methods for molecular generation, and ML models for prediction of material properties, which serve as a tutorial for researchers who have little experience with ML before and want to apply ML for various applications. Last but not least, it draws insights into the current limitations of ML-guided design of organic molecules and polymers. We anticipate that ML-assisted materials design for organic molecules and polymers will be the driving force in the near future, to meet the tremendous demand of new materials with tailored properties in different fields.

Towards accurate high-throughput ligand affinity prediction by exploiting structural ensembles, docking metrics and ligand similarity

MOTIVATION

Nowadays, virtual screening (VS) plays a major role in the process of drug development. Nonetheless, an accurate estimation of binding affinities, which is crucial at all stages, is not trivial and may require target-specific fine-tuning. Furthermore, drug design also requires improved predictions for putative secondary targets among which is Estrogen Receptor alpha (ERα).

RESULTS

VS based on combinations of Structure-Based VS (SBVS) and Ligand-Based VS (LBVS) is gaining momentum to improve VS performances. In this study, we propose an integrated approach using ligand docking on multiple structural ensembles to reflect receptor flexibility. Then, we investigate the impact of the two different types of features (structure-based and ligand molecular descriptors) on affinity predictions using a random forest algorithm. We find that ligand-based features have lower predictive power (rP = 0.69, R2 = 0.47) than structure-based features (rP = 0.78, R2 = 0.60). Their combination maintains high accuracy (rP = 0.73, R2 = 0.50) on the internal test set, but it shows superior robustness on external datasets. Further improvement and extending the training dataset to include xenobiotics, leads to a novel high-throughput affinity prediction method for ERα ligands (rP = 0.85, R2 = 0.71). The presented prediction tool is provided to the community as a dedicated satellite of the @TOME server in which one can upload a ligand dataset in mol2 format and get ligand docked and affinity predicted.

AVAILABILITY AND IMPLEMENTATION

http://edmon.cbs.cnrs.fr.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Interpretable machine learning as a tool for scientific discovery in chemistry

There has been an upsurge of interest in applying machine learning to chemistry, and impressive predictive accuracies have been achieved, but this has been done without providing any insight into what has been learnt from the training data.

Fast-developing machine learning support complex system research in environmental chemistry

Machine learning will radically accelerate analysis of complex material networks in environmental chemistry.

Identification of clinical and urine biomarkers for uncomplicated urinary tract infection using machine learning algorithms

Women with uncomplicated urinary tract infection (UTI) symptoms are commonly treated with empirical antibiotics, resulting in overuse of antibiotics, which promotes antimicrobial resistance. Available diagnostic tools are either not cost-effective or diagnostically sub-optimal. Here, we identified clinical and urinary immunological predictors for UTI diagnosis. We explored 17 clinical and 42 immunological potential predictors for bacterial culture among women with uncomplicated UTI symptoms using random forest or support vector machine coupled with recursive feature elimination. Urine cloudiness was the best performing clinical predictor to rule out (negative likelihood ratio [LR−] = 0.4) and rule in (LR+ = 2.6) UTI. Using a more discriminatory scale to assess cloudiness (turbidity) increased the accuracy of UTI prediction further (LR+ = 4.4). Urinary levels of MMP9, NGAL, CXCL8 and IL-1β together had a higher LR+ (6.1) and similar LR− (0.4), compared to cloudiness. Varying the bacterial count thresholds for urine culture positivity did not alter best clinical predictor selection, but did affect the number of immunological predictors required for reaching an optimal prediction. We conclude that urine cloudiness is particularly helpful in ruling out negative UTI cases. The identified urinary biomarkers could be used to develop a point of care test for UTI but require further validation.

Reprogramming biological peptides to combat infectious diseases

With the rapid spread of resistance among parasites and bacterial pathogens, antibiotic-resistant infections have drawn much attention worldwide. Consequently, there is an urgent need to develop new strategies to treat neglected diseases and drug-resistant infections. Here, we outline several new strategies that have been developed to counter pathogenic microorganisms by designing and constructing antimicrobial peptides (AMPs). In addition to traditional discovery and design mechanisms guided by chemical biology, synthetic biology and computationally-based approaches offer useful tools for the discovery and generation of bioactive peptides. We believe that the convergence of such fields, coupled with systematic experimentation in animal models, will help translate biological peptides into the clinic. The future of anti-infective therapeutics is headed towards specifically designed molecules whose form is driven by computer-based frameworks. These molecules are selective, stable, and active at therapeutic doses.

Concepts and Core Principles of Fragment-Based Drug Design

In this review, a general introduction to fragment-based drug design and the underlying concepts is given. General considerations and methodologies ranging from library selection/construction over biophysical screening and evaluation methods to in-depth hit qualification and subsequent optimization strategies are discussed. These principles can be generally applied to most classes of drug targets. The examples given for fragment growing, merging, and linking strategies at the end of the review are set in the fields of enzyme-inhibitor design and macromolecule–macromolecule interaction inhibition. Building upon the foundation of fragment-based drug discovery (FBDD) and its methodologies, we also highlight a few new trends in FBDD.

Development of Predictive Models for Identifying Potential S100A9 Inhibitors Based on Machine Learning Methods

S100A9 is a potential therapeutic target for various disease including prostate cancer, colorectal cancer, and Alzheimer's disease. However, the sparsity of atomic level data, such as protein-protein interaction of S100A9 with RAGE, TLR4/MD2, or CD147 (EMMPRIN) hinders the rational drug design of S100A9 inhibitors. Herein we first report predictive models of S100A9 inhibitory effect by applying machine learning classifiers on 2D-molecular descriptors. The models were optimized through feature selectors as well as classifiers to produce the top eight random forest models with robust predictability and high cost-effectiveness. Notably, optimal feature sets were obtained after the reduction of 2,798 features into dozens of features with the chopping of fingerprint bits. Moreover, the high efficiency of compact feature sets allowed us to further screen a large-scale dataset (over 6,000,000 compounds) within a week. Through a consensus vote of the top models, 46 hits (hit rate = 0.000713%) were identified as potential S100A9 inhibitors. We expect that our models will facilitate the drug discovery process by providing high predictive power as well as cost-reduction ability and give insights into designing novel drugs targeting S100A9.

Incorporating Explicit Water Molecules and Ligand Conformation Stability in Machine-Learning Scoring Functions

Structure-based drug design is critically dependent on accuracy of molecular docking scoring functions, and there is of significant interest to advance scoring functions with machine learning approaches. In this work, by judiciously expanding the training set, exploring new features related to explicit mediating water molecules as well as ligand conformation stability, and applying extreme gradient boosting (XGBoost) with Δ-Vina parametrization, we have improved robustness and applicability of machine-learning scoring functions. The new scoring function ΔvinaXGB can not only perform consistently among the top compared to classical scoring functions for the CASF-2016 benchmark but also achieves significantly better prediction accuracy in different types of structures that mimic real docking applications.

Advances and Perspectives in Applying Deep Learning for Drug Design and Discovery

Discovering (or planning) a new drug candidate involves many parameters, which makes this process slow, costly, and leading to failures at the end in some cases. In the last decades, we have witnessed a revolution in the computational area (hardware, software, large-scale computing, etc.), as well as an explosion in data generation (big data), which raises the need for more sophisticated algorithms to analyze this myriad of data. In this scenario, we can highlight the potentialities of artificial intelligence (AI) or computational intelligence (CI) as a powerful tool to analyze medicinal chemistry data. According to IEEE, computational intelligence involves the theory, the design, the application, and the development of biologically and linguistically motivated computational paradigms. In addition, CI encompasses three main methodologies: neural networks (NN), fuzzy systems, and evolutionary computation. In particular, artificial neural networks have been successfully applied in medicinal chemistry studies. A branch of the NN area that has attracted a lot of attention refers to deep learning (DL) due to its generalization power and ability to extract features from data. Therefore, in this mini-review we will briefly outline the present scope, advances, and challenges related to the use of DL in drug design and discovery, describing successful studies involving quantitative structure-activity relationships (QSAR) and virtual screening (VS) of databases containing thousands of compounds.

ContraDRG: Automatic Partial Charge Prediction by Machine Learning

In recent years, machine learning techniques have been widely used in biomedical research to predict unseen data based on models trained on experimentally derived data. In the current study, we used machine learning algorithms to emulate computationally complex predictions in a reverse engineering-like manner and developed ContraDRG, a software that can be used to predict partial charges for small molecules based on PRODRG and Automated Topology Builder (ATB) predictions. Both tools generate molecular topology files, including the partial atomic charge, by using different procedures. We show that ContraDRG can accurately predict partial charges in a fraction of the time, because it exploits existing complex models with intensive calculations by using machine learning techniques and thus can also be applied for screening projects with large amounts of molecules. We provide ContraDRG as a web server, which can be used to automatically assign partial charges to incoming user-specified molecules by using our machine learning models. In this study, we compared ContraDRG with PRODRG and ATB in regard of predictivity by statistical methods. ContraDRG allows predicting ATB-derived partial charges with an R2 value up to 0.980 and for PRODRG up to 1.00. While ATB requires hours or days for the quantum mechanical accurate calculation and refinements, ContraDRG does its approximation within seconds.

Fingerprinting CANDO: Increased Accuracy with Structure- and Ligand-Based Shotgun Drug Repurposing

We have upgraded our Computational Analysis of Novel Drug Opportunities (CANDO) platform for shotgun drug repurposing by including ligand-based, data fusion, and decision tree pipelines. The goal of shotgun drug repurposing is to screen and rank every existing human use drug or compound for every disease/indication. The first version of CANDO implemented a structure-based pipeline that modeled interactions between compounds and proteins on a large scale, generating compound-proteome interaction signatures used to infer the similarity of drug behavior; the new pipelines accomplish this by incorporating molecular fingerprints and the Tanimoto coefficient. We obtain improved benchmarking performance with the new pipelines across all three evaluation metrics used: average indication accuracy, pairwise accuracy, and coverage. The best performing pipeline achieves an average indication accuracy of 19.0% at the top10 cutoff, compared to 11.7% for v1, and 2.2% for a random control. Our results demonstrate that the CANDO drug recovery accuracy is substantially improved by integrating multiple pipelines, thereby enhancing our ability to generate putative therapeutic repurposing candidates, and increasing drug discovery efficiency.

Compound optimization monitor (COMO) method for computational evaluation of progress in medicinal chemistry projects

Aim: Development of a new, practically applicable computational method to monitor progress in lead optimization. Computational approaches that aid in compound optimization are discussed and the Compound Optimization Monitor (COMO) method is introduced and put into scientific context. Methodology & calculations: The methodological concept and the COMO scoring scheme are described in detail. Results & discussions: Calculation parameters are evaluated, and profiling results reported for an ensemble of analog series. Future perspective: The dual role of virtual analogs as diagnostic tools for progress evaluation and as potential candidates for lead optimization is discussed. In light of this dual role, interfacing COMO with machine learning for compound activity prediction and prioritization of candidates is highlighted as a future research objective.

OnionNet: a Multiple-Layer Intermolecular-Contact-Based Convolutional Neural Network for Protein-Ligand Binding Affinity Prediction

Computational drug discovery provides an efficient tool for helping large-scale lead molecule screening. One of the major tasks of lead discovery is identifying molecules with promising binding affinities toward a target, a protein in general. The accuracies of current scoring functions that are used to predict the binding affinity are not satisfactory enough. Thus, machine learning or deep learning based methods have been developed recently to improve the scoring functions. In this study, a deep convolutional neural network model (called OnionNet) is introduced; its features are based on rotation-free element-pair-specific contacts between ligands and protein atoms, and the contacts are further grouped into different distance ranges to cover both the local and nonlocal interaction information between the ligand and the protein. The prediction power of the model is evaluated and compared with other scoring functions using the comparative assessment of scoring functions (CASF-2013) benchmark and the v2016 core set of the PDBbind database. The robustness of the model is further explored by predicting the binding affinities of the complexes generated from docking simulations instead of experimentally determined PDB structures.

Interpretation of Compound Activity Predictions from Complex Machine Learning Models Using Local Approximations and Shapley Values

In qualitative or quantitative studies of structure-activity relationships (SARs), machine learning (ML) models are trained to recognize structural patterns that differentiate between active and inactive compounds. Understanding model decisions is challenging but of critical importance to guide compound design. Moreover, the interpretation of ML results provides an additional level of model validation based on expert knowledge. A number of complex ML approaches, especially deep learning (DL) architectures, have distinctive black-box character. Herein, a locally interpretable explanatory method termed Shapley additive explanations (SHAP) is introduced for rationalizing activity predictions of any ML algorithm, regardless of its complexity. Models resulting from random forest (RF), nonlinear support vector machine (SVM), and deep neural network (DNN) learning are interpreted, and structural patterns determining the predicted probability of activity are identified and mapped onto test compounds. The results indicate that SHAP has high potential for rationalizing predictions of complex ML models.

All-Assay-Max2 pQSAR: Activity Predictions as Accurate as Four-Concentration IC50s for 8558 Novartis Assays

Profile-quantitative structure-activity relationship (pQSAR) is a massively multitask, two-step machine learning method with unprecedented scope, accuracy, and applicability domain. In step one, a "profile" of conventional single-assay random forest regression models are trained on a very large number of biochemical and cellular pIC₅₀ assays using Morgan 2 substructural fingerprints as compound descriptors. In step two, a panel of partial least squares (PLS) models are built using the profile of pIC₅₀ predictions from those random forest regression models as compound descriptors (hence the name). Previously described for a panel of 728 biochemical and cellular kinase assays, we have now built an enormous pQSAR from 11 805 diverse Novartis (NVS) IC₅₀ and EC₅₀ assays. This large number of assays, and hence of compound descriptors for PLS, dictated reducing the profile by only including random forest regression models whose predictions correlate with the assay being modeled. The random forest regression and pQSAR models were evaluated with our "realistically novel" held-out test set, whose median average similarity to the nearest training set member across the 11 805 assays was only 0.34, comparable to the novelty of compounds actually selected from virtual screens. For the 11 805 single-assay random forest regression models, the median correlation of prediction with the experiment was only r_ext² = 0.05, virtually random, and only 8% of the models achieved our standard success threshold of r_ext² = 0.30. For pQSAR, the median correlation was r_ext² = 0.53, comparable to four-concentration experimental IC₅₀s, and 72% of the models met our r_ext² > 0.30 standard, totaling 8558 successful models. The successful models included assays from all of the 51 annotated target subclasses, as well as 4196 phenotypic assays, indicating that pQSAR can be applied to virtually any disease area. Every month, all models are updated to include new measurements, and predictions are made for 5.5 million NVS compounds, totaling 50 billion predictions. Common uses have included virtual screening, selectivity design, toxicity and promiscuity prediction, mechanism-of-action prediction, and others. Several such actual applications are described.

A general approach for retrosynthetic molecular core analysis

Scaffold analysis of compound data sets has reemerged as a chemically interpretable alternative to machine learning for chemical space and structure–activity relationships analysis. In this context, analog series-based scaffolds (ASBS) are synthetically relevant core structures that represent individual series of analogs. As an extension to ASBS, we herein introduce the development of a general conceptual framework that considers all putative cores of molecules in a compound data set, thus softening the often applied “single molecule–single scaffold” correspondence. A putative core is here defined as any substructure of a molecule complying with two basic rules: (a) the size of the core is a significant proportion of the whole molecule size and (b) the substructure can be reached from the original molecule through a succession of retrosynthesis rules. Thereafter, a bipartite network consisting of molecules and cores can be constructed for a database of chemical structures. Compounds linked to the same cores are considered analogs. We present case studies illustrating the potential of the general framework. The applications range from inter- and intra-core diversity analysis of compound data sets, structure–property relationships, and identification of analog series and ASBS. The molecule–core network herein presented is a general methodology with multiple applications in scaffold analysis. New statistical methods are envisioned that will be able to draw quantitative conclusions from these data. The code to use the method presented in this work is freely available as an additional file. Follow-up applications include analog searching and core structure–property relationships analyses.

Molecular Docking: Shifting Paradigms in Drug Discovery

Molecular docking is an established in silico structure-based method widely used in drug discovery. Docking enables the identification of novel compounds of therapeutic interest, predicting ligand-target interactions at a molecular level, or delineating structure-activity relationships (SAR), without knowing a priori the chemical structure of other target modulators. Although it was originally developed to help understanding the mechanisms of molecular recognition between small and large molecules, uses and applications of docking in drug discovery have heavily changed over the last years. In this review, we describe how molecular docking was firstly applied to assist in drug discovery tasks. Then, we illustrate newer and emergent uses and applications of docking, including prediction of adverse effects, polypharmacology, drug repurposing, and target fishing and profiling, discussing also future applications and further potential of this technique when combined with emergent techniques, such as artificial intelligence.

A Drug Decision Support System for Developing a Successful Drug Candidate Using Machine Learning Techniques

BACKGROUND

Virtual screening of candidate drug molecules using machine learning techniques plays a key role in pharmaceutical industry to design and discovery of new drugs. Computational classification methods can determine drug types according to the disease groups and distinguish approved drugs from withdrawn ones.

INTRODUCTION

Classification models developed in this study can be used as a simple filter in drug modelling to eliminate potentially inappropriate molecules in the early stages. In this work, we developed a Drug Decision Support System (DDSS) to classify each drug candidate molecule as potentially drug or non-drug and to predict its disease group.

METHODS

Molecular descriptors were identified for the determination of a number of rules in drug molecules. They were derived using ADRIANA.Code program and Lipinski's rule of five. We used Artificial Neural Network (ANN) to classify drug molecules correctly according to the types of diseases. Closed frequent molecular structures in the form of subgraph fragments were also obtained with Gaston algorithm included in ParMol Package to find common molecular fragments for withdrawn drugs.

RESULTS

We observed that TPSA, XlogP Natoms, HDon_O and TPSA are the most distinctive features in the pool of the molecular descriptors and evaluated the performances of classifiers on all datasets and found that classification accuracies are very high on all the datasets. Neural network models achieved 84.6% and 83.3% accuracies on test sets including cardiac therapy, anti-epileptics and anti-parkinson drugs with approved and withdrawn drugs for drug classification problems.

CONCLUSION

The experimental evaluation shows that the system is promising at determination of potential drug molecules to classify drug molecules correctly according to the types of diseases.

The Aristotle Classifier: Using the Whole Glycomic Profile To Indicate a Disease State

"The totality is not, as it were, a mere heap, but the whole is something besides the parts."-Aristotle. We built a classifier that uses the totality of the glycomic profile, not restricted to a few glycoforms, to differentiate samples from two different sources. This approach, which relies on using thousands of features, is a radical departure from current strategies, where most of the glycomic profile is ignored in favor of selecting a few features, or even a single feature, meant to capture the differences in sample types. The classifier can be used to differentiate the source of the material; applicable sources may be different species of animals, different protein production methods, or, most importantly, different biological states (disease vs healthy). The classifier can be used on glycomic data in any form, including derivatized monosaccharides, intact glycans, or glycopeptides. It takes advantage of the fact that changing the source material can cause a change in the glycomic profile in many subtle ways: some glycoforms can be upregulated, some downregulated, some may appear unchanged, yet their proportion-with respect to other forms present-can be altered to a detectable degree. By classifying samples using the entirety of their glycan abundances, along with the glycans' relative proportions to each other, the "Aristotle Classifier" is more effective at capturing the underlying trends than standard classification procedures used in glycomics, including PCA (principal components analysis). It also outperforms workflows where a single, representative glycomic-based biomarker is used to classify samples. We describe the Aristotle Classifier and provide several examples of its utility for biomarker studies and other classification problems using glycomic data from several sources.

Systematic selection of chemical fingerprint features improves the Gibbs energy prediction of biochemical reactions

MOTIVATION

Accurate and wide-ranging prediction of thermodynamic parameters for biochemical reactions can facilitate deeper insights into the workings and the design of metabolic systems.

RESULTS

Here, we introduce a machine learning method with chemical fingerprint-based features for the prediction of the Gibbs free energy of biochemical reactions. From a large pool of 2D fingerprint-based features, this method systematically selects a small number of relevant ones and uses them to construct a regularized linear model. Since a manual selection of 2D structure-based features can be a tedious and time-consuming task, requiring expert knowledge about the structure-activity relationship of chemical compounds, the systematic feature selection step in our method offers a convenient means to identify relevant 2D fingerprint-based features. By comparing our method with state-of-the-art linear regression-based methods for the standard Gibbs free energy prediction, we demonstrated that its prediction accuracy and prediction coverage are most favorable. Our results show direct evidence that a number of 2D fingerprints collectively provide useful information about the Gibbs free energy of biochemical reactions and that our systematic feature selection procedure provides a convenient way to identify them.

AVAILABILITY AND IMPLEMENTATION

Our software is freely available for download at http://sfb.kaust.edu.sa/Pages/Software.aspx.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Improving prediction of phenotypic drug response on cancer cell lines using deep convolutional network

Background

Understanding the phenotypic drug response on cancer cell lines plays a vital role in anti-cancer drug discovery and re-purposing. The Genomics of Drug Sensitivity in Cancer (GDSC) database provides open data for researchers in phenotypic screening to build and test their models. Previously, most research in these areas starts from the molecular fingerprints or physiochemical features of drugs, instead of their structures.

Results

In this paper, a model called twin Convolutional Neural Network for drugs in SMILES format (tCNNS) is introduced for phenotypic screening. tCNNS uses a convolutional network to extract features for drugs from their simplified molecular input line entry specification (SMILES) format and uses another convolutional network to extract features for cancer cell lines from the genetic feature vectors respectively. After that, a fully connected network is used to predict the interaction between the drugs and the cancer cell lines. When the training set and the testing set are divided based on the interaction pairs between drugs and cell lines, tCNNS achieves 0.826, 0.831 for the mean and top quartile of the coefficient of determinant (R²) respectively and 0.909, 0.912 for the mean and top quartile of the Pearson correlation (R_p) respectively, which are significantly better than those of the previous works (Ammad-Ud-Din et al., J Chem Inf Model 54:2347–9, 2014), (Haider et al., PLoS ONE 10:0144490, 2015), (Menden et al., PLoS ONE 8:61318, 2013). However, when the training set and the testing set are divided exclusively based on drugs or cell lines, the performance of tCNNS decreases significantly and R_p and R² drop to barely above 0.

Conclusions

Our approach is able to predict the drug effects on cancer cell lines with high accuracy, and its performance remains stable with less but high-quality data, and with fewer features for the cancer cell lines. tCNNS can also solve the problem of outliers in other feature space. Besides achieving high scores in these statistical metrics, tCNNS also provides some insights into the phenotypic screening. However, the performance of tCNNS drops in the blind test.

Electronic supplementary material

The online version of this article (10.1186/s12859-019-2910-6) contains supplementary material, which is available to authorized users.

Targeting HIV/HCV Coinfection Using a Machine Learning-Based Multiple Quantitative Structure-Activity Relationships (Multiple QSAR) Method

Human immunodeficiency virus type-1 and hepatitis C virus (HIV/HCV) coinfection occurs when a patient is simultaneously infected with both human immunodeficiency virus type-1 (HIV-1) and hepatitis C virus (HCV), which is common today in certain populations. However, the treatment of coinfection is a challenge because of the special considerations needed to ensure hepatic safety and avoid drug–drug interactions. Multitarget inhibitors with less toxicity may provide a promising therapeutic strategy for HIV/HCV coinfection. However, the identification of one molecule that acts on multiple targets simultaneously by experimental evaluation is costly and time-consuming. In silico target prediction tools provide more opportunities for the development of multitarget inhibitors. In this study, by combining Naïve Bayes (NB) and support vector machine (SVM) algorithms with two types of molecular fingerprints, MACCS and extended connectivity fingerprints 6 (ECFP6), 60 classification models were constructed to predict compounds that were active against 11 HIV-1 targets and four HCV targets based on a multiple quantitative structure–activity relationships (multiple QSAR) method. Five-fold cross-validation and test set validation were performed to measure the performance of the 60 classification models. Our results show that the 60 multiple QSAR models appeared to have high classification accuracy in terms of the area under the ROC curve (AUC) values, which ranged from 0.83 to 1 with a mean value of 0.97 for the HIV-1 models and from 0.84 to 1 with a mean value of 0.96 for the HCV models. Furthermore, the 60 models were used to comprehensively predict the potential targets of an additional 46 compounds, including 27 approved HIV-1 drugs, 10 approved HCV drugs and nine selected compounds known to be active against one or more targets of HIV-1 or HCV. Finally, 20 hits, including seven approved HIV-1 drugs, four approved HCV drugs, and nine other compounds, were predicted to be HIV/HCV coinfection multitarget inhibitors. The reported bioactivity data confirmed that seven out of nine compounds actually interacted with HIV-1 and HCV targets simultaneously with diverse binding affinities. The remaining predicted hits and chemical-protein interaction pairs with the potential ability to suppress HIV/HCV coinfection are worthy of further experimental investigation. This investigation shows that the multiple QSAR method is useful in predicting chemical-protein interactions for the discovery of multitarget inhibitors and provides a unique strategy for the treatment of HIV/HCV coinfection.

Deep Reinforcement Learning for Multiparameter Optimization in de novo Drug Design

In medicinal chemistry programs it is key to design and make compounds that are efficacious and safe. This is a long, complex, and difficult multiparameter optimization process, often including several properties with orthogonal trends. New methods for the automated design of compounds against profiles of multiple properties are thus of great value. Here we present a fragment-based reinforcement learning approach based on an actor-critic model, for the generation of novel molecules with optimal properties. The actor and the critic are both modeled with bidirectional long short-term memory (LSTM) networks. The AI method learns how to generate new compounds with desired properties by starting from an initial set of lead molecules and then improving these by replacing some of their fragments. A balanced binary tree based on the similarity of fragments is used in the generative process to bias the output toward structurally similar molecules. The method is demonstrated by a case study showing that 93% of the generated molecules are chemically valid and more than a third satisfy the targeted objectives, while there were none in the initial set.

The Application of Machine Learning Techniques in Clinical Drug Therapy

INTRODUCTION

The development of a novel drug is an extremely complicated process that includes the target identification, design and manufacture, and proper therapy of the novel drug, as well as drug dose selection, drug efficacy evaluation, and adverse drug reaction control. Due to the limited resources, high costs, long duration, and low hit-to-lead ratio in the development of pharmacogenetics and computer technology, machine learning techniques have assisted novel drug development and have gradually received more attention by researchers.

METHODS

According to current research, machine learning techniques are widely applied in the process of the discovery of new drugs and novel drug targets, the decision surrounding proper therapy and drug dose, and the prediction of drug efficacy and adverse drug reactions.

RESULTS AND CONCLUSION

In this article, we discussed the history, workflow, and advantages and disadvantages of machine learning techniques in the processes mentioned above. Although the advantages of machine learning techniques are fairly obvious, the application of machine learning techniques is currently limited. With further research, the application of machine techniques in drug development could be much more widespread and could potentially be one of the major methods used in drug development.

QSAR Classification-Based Virtual Screening Followed by Molecular Docking Identification of Potential COX-2 Inhibitors in a Natural Product Library

Developments of natural inhibitors to prevent the function of cyclooxygenase-2 (COX-2) protein, responsible for a variety of inflammations and cancers, are a major challenge in the scientific community. In this study, robust QSAR classification models for predicting COX-2 inhibitor were developed, by which the self-organizing feature map neural network and random forest (RF) were adopted to improve the prediction of classification model ability. The F-score-based criterion combined with RF was used for feature selection, and good performance for COX-2 inhibitor prediction in overall accuracy was demonstrated. We used this model as a virtual screening tool for identifying the potential COX-2 inhibitor from a natural product library and found potential hit compounds. This compound further screened by applying molecular docking simulation identified five potential hits such as osthole, kavain, vanillyl acetone, myristicin, and psoralen, having a comparable binding affinity to COX-2 protein. However, in cell experiment, three hit compounds revealed COX-2 inhibitory activity in mRNA and protein level such as osthole, kavain, and psoralen.