Accurate and efficient target prediction using a potency-sensitive influence-relevance voter

Quantum Mechanics and Machine Learning Synergies: Graph Attention Neural Networks to Predict Chemical Reactivity

There is a lack of scalable quantitative measures of reactivity that cover the full range of functional groups in organic chemistry, ranging from highly unreactive C-C bonds to highly reactive naked ions. Measuring reactivity experimentally is costly and time-consuming, and no single method has sufficient dynamic range to cover the astronomical size of chemical reactivity space. In previous quantum chemistry studies, we have introduced Methyl Cation Affinities (MCA*) and Methyl Anion Affinities (MAA*), using a solvation model, as quantitative measures of reactivity for organic functional groups over the broadest range. Although MCA* and MAA* offer good estimates of reactivity parameters, their calculation through Density Functional Theory (DFT) simulations is time-consuming. To circumvent this problem, we first use DFT to calculate MCA* and MAA* for more than 2,400 organic molecules thereby establishing a large data set of chemical reactivity scores. We then design deep learning methods to predict the reactivity of molecular structures and train them using this curated data set in combination with different representations of molecular structures. Using 10-fold cross-validation, we show that graph attention neural networks applied to a relational model of molecular structures produce the most accurate estimates of reactivity, achieving over 91% test accuracy for predicting the MCA* ± 3.0 or MAA* ± 3.0, over 50 orders of magnitude. Finally, we demonstrate the application of these reactivity scores to two tasks: (1) chemical reaction prediction and (2) combinatorial generation of reaction mechanisms. The curated data sets of MCA* and MAA* scores is available through the ChemDB chemoinformatics web portal at cdb.ics.uci.edu under Chemical Reactivities data sets.

Adding Stochastic Negative Examples into Machine Learning Improves Molecular Bioactivity Prediction

Multitask deep neural networks learn to predict ligand-target binding by example, yet public pharmacological data sets are sparse, imbalanced, and approximate. We constructed two hold-out benchmarks to approximate temporal and drug-screening test scenarios, whose characteristics differ from a random split of conventional training data sets. We developed a pharmacological data set augmentation procedure, Stochastic Negative Addition (SNA), which randomly assigns untested molecule-target pairs as transient negative examples during training. Under the SNA procedure, drug-screening benchmark performance increases from R² = 0.1926 ± 0.0186 to 0.4269 ± 0.0272 (122%). This gain was accompanied by a modest decrease in the temporal benchmark (13%). SNA increases in drug-screening performance were consistent for classification and regression tasks and outperformed y-randomized controls. Our results highlight where data and feature uncertainty may be problematic and how leveraging uncertainty into training improves predictions of drug-target relationships.

Deep Learning for Drug Discovery and Cancer Research: Automated Analysis of Vascularization Images

Likely drug candidates which are identified in traditional pre-clinical drug screens often fail in patient trials, increasing the societal burden of drug discovery. A major contributing factor to this phenomenon is the failure of traditional in vitro models of drug response to accurately mimic many of the more complex properties of human biology. We have recently introduced a new microphysiological system for growing vascularized, perfused microtissues that more accurately models human physiology and is suitable for large drug screens. In this work, we develop a machine learning model that can quickly and accurately flag compounds which effectively disrupt vascular networks from images taken before and after drug application in vitro. The system is based on a convolutional neural network and achieves near perfect accuracy while committing potentially no expensive false negatives.

Web-Based Tools for Polypharmacology Prediction

Drug promiscuity or polypharmacology is the ability of small molecules to interact with multiple protein targets simultaneously. In drug discovery, understanding the polypharmacology of potential drug molecules is crucial to improve their efficacy and safety, and to discover the new therapeutic potentials of existing drugs. Over the past decade, several computational methods have been developed to study the polypharmacology of small molecules, many of which are available as Web services. In this chapter, we review some of these Web tools focusing on ligand based approaches. We highlight in particular our recently developed polypharmacology browser (PPB) and its application for finding the side targets of a new inhibitor of the TRPV6 calcium channel.

Polypharmacology Browser PPB2: Target Prediction Combining Nearest Neighbors with Machine Learning

Here we report PPB2 as a target prediction tool assigning targets to a query molecule based on ChEMBL data. PPB2 computes ligand similarities using molecular fingerprints encoding composition (MQN), molecular shape and pharmacophores (Xfp), and substructures (ECfp4) and features an unprecedented combination of nearest neighbor (NN) searches and Naı̈ve Bayes (NB) machine learning, together with simple NN searches, NB and Deep Neural Network (DNN) machine learning models as further options. Although NN(ECfp4) gives the best results in terms of recall in a 10-fold cross-validation study, combining NN searches with NB machine learning provides superior precision statistics, as well as better results in a case study predicting off-targets of a recently reported TRPV6 calcium channel inhibitor, illustrating the value of this combined approach. PPB2 is available to assess possible off-targets of small molecule drug-like compounds by public access at http://gdb.unibe.ch .

Practical Model Selection for Prospective Virtual Screening

Virtual (computational) high-throughput screening provides a strategy for prioritizing compounds for experimental screens, but the choice of virtual screening algorithm depends on the data set and evaluation strategy. We consider a wide range of ligand-based machine learning and docking-based approaches for virtual screening on two protein–protein interactions, PriA-SSB and RMI-FANCM, and present a strategy for choosing which algorithm is best for prospective compound prioritization. Our workflow identifies a random forest as the best algorithm for these targets over more sophisticated neural network-based models. The top 250 predictions from our selected random forest recover 37 of the 54 active compounds from a library of 22,434 new molecules assayed on PriA-SSB. We show that virtual screening methods that perform well on public data sets and synthetic benchmarks, like multi-task neural networks, may not always translate to prospective screening performance on a specific assay of interest.

Finding the molecular scaffold of nuclear receptor inhibitors through high-throughput screening based on proteochemometric modelling

Nuclear receptors (NR) are a class of proteins that are responsible for sensing steroid and thyroid hormones and certain other molecules. In that case, NR have the ability to regulate the expression of specific genes and associated with various diseases, which make it essential drug targets. Approaches which can predict the inhibition ability of compounds for different NR target should be particularly helpful for drug development. In this study, proteochemometric modelling was introduced to analysis the bioactivity between chemical compounds and NR targets. Results illustrated the ability of our PCM model for high-throughput NR-inhibitor screening after evaluated on both internal (AUC > 0.870) and external (AUC > 0.746) validation set. Moreover, in-silico predicted bioactive compounds were clustered according to structure similarity and a series of representative molecular scaffolds can be derived for five major NR targets. Through scaffolds analysis, those essential bioactive scaffolds of different NR target can be detected and compared. Generally, the methods and molecular scaffolds proposed in this article can not only help the screening of potential therapeutic NR-inhibitors but also able to guide the future NR-related drug discovery.

Opportunities and obstacles for deep learning in biology and medicine

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems-patient classification, fundamental biological processes and treatment of patients-and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural network's prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine.

Large-Scale Prediction of Drug-Target Interaction: a Data-Centric Review

The prediction of drug-target interactions (DTIs) is of extraordinary significance to modern drug discovery in terms of suggesting new drug candidates and repositioning old drugs. Despite technological advances, large-scale experimental determination of DTIs is still expensive and laborious. Effective and low-cost computational alternatives remain in strong need. Meanwhile, open-access resources have been rapidly growing with massive amount of bioactivity data becoming available, creating unprecedented opportunities for the development of novel in silico models for large-scale DTI prediction. In this work, we review the state-of-the-art computational approaches for identifying DTIs from a data-centric perspective: what the underlying data are and how they are utilized in each study. We also summarize popular public data resources and online tools for DTI prediction. It is found that various types of data were employed including properties of chemical structures, drug therapeutic effects and side effects, drug-target binding, drug-drug interactions, bioactivity data of drug molecules across multiple biological targets, and drug-induced gene expressions. More often, the heterogeneous data were integrated to offer better performance. However, challenges remain such as handling data imbalance, incorporating negative samples and quantitative bioactivity data, as well as maintaining cross-links among different data sources, which are essential for large-scale and automated information integration.

Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set

The increase of publicly available bioactivity data in recent years has fueled and catalyzed research in chemogenomics, data mining, and modeling approaches. As a direct result, over the past few years a multitude of different methods have been reported and evaluated, such as target fishing, nearest neighbor similarity-based methods, and Quantitative Structure Activity Relationship (QSAR)-based protocols. However, such studies are typically conducted on different datasets, using different validation strategies, and different metrics. In this study, different methods were compared using one single standardized dataset obtained from ChEMBL, which is made available to the public, using standardized metrics (BEDROC and Matthews Correlation Coefficient). Specifically, the performance of Naïve Bayes, Random Forests, Support Vector Machines, Logistic Regression, and Deep Neural Networks was assessed using QSAR and proteochemometric (PCM) methods. All methods were validated using both a random split validation and a temporal validation, with the latter being a more realistic benchmark of expected prospective execution. Deep Neural Networks are the top performing classifiers, highlighting the added value of Deep Neural Networks over other more conventional methods. Moreover, the best method ('DNN_PCM') performed significantly better at almost one standard deviation higher than the mean performance. Furthermore, Multi-task and PCM implementations were shown to improve performance over single task Deep Neural Networks. Conversely, target prediction performed almost two standard deviations under the mean performance. Random Forests, Support Vector Machines, and Logistic Regression performed around mean performance. Finally, using an ensemble of DNNs, alongside additional tuning, enhanced the relative performance by another 27% (compared with unoptimized 'DNN_PCM'). Here, a standardized set to test and evaluate different machine learning algorithms in the context of multi-task learning is offered by providing the data and the protocols. Graphical Abstract .

Shallow Representation Learning via Kernel PCA Improves QSAR Modelability

Linear models offer a robust, flexible, and computationally efficient set of tools for modeling quantitative structure-activity relationships (QSARs) but have been eclipsed in performance by nonlinear methods. Support vector machines (SVMs) and neural networks are currently among the most popular and accurate QSAR methods because they learn new representations of the data that greatly improve modelability. In this work, we use shallow representation learning to improve the accuracy of L1 regularized logistic regression (LASSO) and meet the performance of Tanimoto SVM. We embedded chemical fingerprints in Euclidean space using Tanimoto (a.k.a. Jaccard) similarity kernel principal component analysis (KPCA) and compared the effects on LASSO and SVM model performance for predicting the binding activities of chemical compounds against 102 virtual screening targets. We observed similar performance and patterns of improvement for LASSO and SVM. We also empirically measured model training and cross-validation times to show that KPCA used in concert with LASSO classification is significantly faster than linear SVM over a wide range of training set sizes. Our work shows that powerful linear QSAR methods can match nonlinear methods and demonstrates a modular approach to nonlinear classification that greatly enhances QSAR model prototyping facility, flexibility, and transferability.

Deep learning for computational chemistry

The rise and fall of artificial neural networks is well documented in the scientific literature of both computer science and computational chemistry. Yet almost two decades later, we are now seeing a resurgence of interest in deep learning, a machine learning algorithm based on multilayer neural networks. Within the last few years, we have seen the transformative impact of deep learning in many domains, particularly in speech recognition and computer vision, to the extent that the majority of expert practitioners in those field are now regularly eschewing prior established models in favor of deep learning models. In this review, we provide an introductory overview into the theory of deep neural networks and their unique properties that distinguish them from traditional machine learning algorithms used in cheminformatics. By providing an overview of the variety of emerging applications of deep neural networks, we highlight its ubiquity and broad applicability to a wide range of challenges in the field, including quantitative structure activity relationship, virtual screening, protein structure prediction, quantum chemistry, materials design, and property prediction. In reviewing the performance of deep neural networks, we observed a consistent outperformance against non-neural networks state-of-the-art models across disparate research topics, and deep neural network-based models often exceeded the "glass ceiling" expectations of their respective tasks. Coupled with the maturity of GPU-accelerated computing for training deep neural networks and the exponential growth of chemical data on which to train these networks on, we anticipate that deep learning algorithms will be a valuable tool for computational chemistry. © 2017 Wiley Periodicals, Inc.

ExCAPE-DB: an integrated large scale dataset facilitating Big Data analysis in chemogenomics

Chemogenomics data generally refers to the activity data of chemical compounds on an array of protein targets and represents an important source of information for building in silico target prediction models. The increasing volume of chemogenomics data offers exciting opportunities to build models based on Big Data. Preparing a high quality data set is a vital step in realizing this goal and this work aims to compile such a comprehensive chemogenomics dataset. This dataset comprises over 70 million SAR data points from publicly available databases (PubChem and ChEMBL) including structure, target information and activity annotations. Our aspiration is to create a useful chemogenomics resource reflecting industry-scale data not only for building predictive models of in silico polypharmacology and off-target effects but also for the validation of cheminformatics approaches in general.

The polypharmacology browser: a web-based multi-fingerprint target prediction tool using ChEMBL bioactivity data

Background

Several web-based tools have been reported recently which predict the possible targets of a small molecule by similarity to compounds of known bioactivity using molecular fingerprints (fps), however predictions in each case rely on similarities computed from only one or two fps. Considering that structural similarity and therefore the predicted targets strongly depend on the method used for comparison, it would be highly desirable to predict targets using a broader set of fps simultaneously.

Results

Herein, we present the polypharmacology browser (PPB), a web-based platform which predicts possible targets for small molecules by searching for nearest neighbors using ten different fps describing composition, substructures, molecular shape and pharmacophores. PPB searches through 4613 groups of at least 10 same target annotated bioactive molecules from ChEMBL and returns a list of predicted targets ranked by consensus voting scheme and p value. A validation study across 670 drugs with up to 20 targets showed that combining the predictions from all 10 fps gives the best results, with on average 50% of the known targets of a drug being correctly predicted with a hit rate of 25%. Furthermore, when profiling a new inhibitor of the calcium channel TRPV6 against 24 targets taken from a safety screen panel, we observed inhibition in 5 out of 5 targets predicted by PPB and in 7 out of 18 targets not predicted by PPB. The rate of correct (5/12) and incorrect (0/12) predictions for this compound by PPB was comparable to that of other web-based prediction tools.

Conclusion

PPB offers a versatile platform for target prediction based on multi-fingerprint comparisons, and is freely accessible at www.gdb.unibe.ch as a valuable support for drug discovery.Graphical abstract

bioassayR: Cross-Target Analysis of Small Molecule Bioactivity

Despite a large and rapidly growing body of small molecule bioactivity screens available in the public domain, systematic leverage of the data to assess target druggability and compound selectivity has been confounded by a lack of suitable cross-target analysis software. We have developed bioassayR, a computational tool that enables simultaneous analysis of thousands of bioassay experiments performed over a diverse set of compounds and biological targets. Unique features include support for large-scale cross-target analyses of both public and custom bioassays, generation of high throughput screening fingerprints (HTSFPs), and an optional preloaded database that provides access to a substantial portion of publicly available bioactivity data. bioassayR is implemented as an open-source R/Bioconductor package available from https://bioconductor.org/packages/bioassayR/.