Target-driven machine learning-enabled virtual screening (TAME-VS) platform for early-stage hit identification.

High-throughput screening (HTS) methods enable the empirical evaluation of a large scale of compounds and can be augmented by virtual screening (VS) techniques to save time and money by using potential active compounds for experimental testing. Structure-based and ligand-based virtual screening approaches have been extensively studied and applied in drug discovery practice with proven outcomes in advancing candidate molecules. However, the experimental data required for VS are expensive, and hit identification in an effective and efficient manner is particularly challenging during early-stage drug discovery for novel protein targets. Herein, we present our TArget-driven Machine learning-Enabled VS (TAME-VS) platform, which leverages existing chemical databases of bioactive molecules to modularly facilitate hit finding. Our methodology enables bespoke hit identification campaigns through a user-defined protein target. The input target ID is used to perform a homology-based target expansion, followed by compound retrieval from a large compilation of molecules with experimentally validated activity. Compounds are subsequently vectorized and adopted for machine learning (ML) model training. These machine learning models are deployed to perform model-based inferential virtual screening, and compounds are nominated based on predicted activity. Our platform was retrospectively validated across ten diverse protein targets and demonstrated clear predictive power. The implemented methodology provides a flexible and efficient approach that is accessible to a wide range of users. The TAME-VS platform is publicly available at https://github.com/bymgood/Target-driven-ML-enabled-VS to facilitate early-stage hit identification.

Structure-function analysis of the SHOC2-MRAS-PP1C holophosphatase complex.

Receptor tyrosine kinase (RTK)-RAS signalling through the downstream mitogen-activated protein kinase (MAPK) cascade regulates cell proliferation and survival. The SHOC2-MRAS-PP1C holophosphatase complex functions as a key regulator of RTK-RAS signalling by removing an inhibitory phosphorylation event on the RAF family of proteins to potentiate MAPK signalling1. SHOC2 forms a ternary complex with MRAS and PP1C, and human germline gain-of-function mutations in this complex result in congenital RASopathy syndromes2-5. However, the structure and assembly of this complex are poorly understood. Here we use cryo-electron microscopy to resolve the structure of the SHOC2-MRAS-PP1C complex. We define the biophysical principles of holoenzyme interactions, elucidate the assembly order of the complex, and systematically interrogate the functional consequence of nearly all of the possible missense variants of SHOC2 through deep mutational scanning. We show that SHOC2 binds PP1C and MRAS through the concave surface of the leucine-rich repeat region and further engages PP1C through the N-terminal disordered region that contains a cryptic RVXF motif. Complex formation is initially mediated by interactions between SHOC2 and PP1C and is stabilized by the binding of GTP-loaded MRAS. These observations explain how mutant versions of SHOC2 in RASopathies and cancer stabilize the interactions of complex members to enhance holophosphatase activity. Together, this integrative structure-function model comprehensively defines key binding interactions within the SHOC2-MRAS-PP1C holophosphatase complex and will inform therapeutic development .

Artificial Intelligent Deep Learning Molecular Generative Modeling of Scaffold-Focused and Cannabinoid CB2 Target-Specific Small-Molecule Sublibraries.

Design and generation of high-quality target- and scaffold-specific small molecules is an important strategy for the discovery of unique and potent bioactive drug molecules. To achieve this goal, authors have developed the deep-learning molecule generation model (DeepMGM) and applied it for the de novo molecular generation of scaffold-focused small-molecule libraries. In this study, a recurrent neural network (RNN) using long short-term memory (LSTM) units was trained with drug-like molecules to result in a general model (g-DeepMGM). Sampling practices on indole and purine scaffolds illustrate the feasibility of creating scaffold-focused chemical libraries based on machine intelligence. Subsequently, a target-specific model (t-DeepMGM) for cannabinoid receptor 2 (CB2) was constructed following the transfer learning process of known CB2 ligands. Sampling outcomes can present similar properties to the reported active molecules. Finally, a discriminator was trained and attached to the DeepMGM to result in an in silico molecular design-test circle. Medicinal chemistry synthesis and biological validation was performed to further investigate the generation outcome, showing that XIE9137 was identified as a potential allosteric modulator of CB2. This study demonstrates how recent progress in deep learning intelligence can benefit drug discovery, especially in de novo molecular design and chemical library generation.

Integrated Multi-Class Classification and Prediction of GPCR Allosteric Modulators by Machine Learning Intelligence.

G-protein-coupled receptors (GPCRs) are the largest and most diverse group of cell surface receptors that respond to various extracellular signals. The allosteric modulation of GPCRs has emerged in recent years as a promising approach for developing target-selective therapies. Moreover, the discovery of new GPCR allosteric modulators can greatly benefit the further understanding of GPCR cell signaling mechanisms. It is critical but also challenging to make an accurate distinction of modulators for different GPCR groups in an efficient and effective manner. In this study, we focus on an 11-class classification task with 10 GPCR subtype classes and a random compounds class. We used a dataset containing 34,434 compounds with allosteric modulators collected from classical GPCR families A, B, and C, as well as random drug-like compounds. Six types of machine learning models, including support vector machine, naïve Bayes, decision tree, random forest, logistic regression, and multilayer perceptron, were trained using different combinations of features including molecular descriptors, Atom-pair fingerprints, MACCS fingerprints, and ECFP6 fingerprints. The performances of trained machine learning models with different feature combinations were closely investigated and discussed. To the best of our knowledge, this is the first work on the multi-class classification of GPCR allosteric modulators. We believe that the classification models developed in this study can be used as simple and accurate tools for the discovery and development of GPCR allosteric modulators.

Generative chemistry: drug discovery with deep learning generative models.

The de novo design of molecular structures using deep learning generative models introduces an encouraging solution to drug discovery in the face of the continuously increased cost of new drug development. From the generation of original texts, images, and videos, to the scratching of novel molecular structures the creativity of deep learning generative models exhibits the height machine intelligence can achieve. The purpose of this paper is to review the latest advances in generative chemistry which relies on generative modeling to expedite the drug discovery process. This review starts with a brief history of artificial intelligence in drug discovery to outline this emerging paradigm. Commonly used chemical databases, molecular representations, and tools in cheminformatics and machine learning are covered as the infrastructure for generative chemistry. The detailed discussions on utilizing cutting-edge generative architectures, including recurrent neural network, variational autoencoder, adversarial autoencoder, and generative adversarial network for compound generation are focused. Challenges and future perspectives follow.

Covalent allosteric modulation: An emerging strategy for GPCRs drug discovery.

Designing covalent allosteric modulators brings new opportunities to the field of drug discovery towards G-protein-coupled receptors (GPCRs). Targeting an allosteric binding pocket can allow a modulator to have protein subtype selectivity and low drug resistance. Utilizing covalent warheads further enables the modulator to increase the binding potency and extend the duration of action. This review starts with GPCR allosteric modulation to discuss the structural biology of allosteric binding pockets, the different types of allosteric modulators, as well as the advantages of employing allosteric modulation. This is followed by a discussion on covalent modulators to clarify how covalent ligands can benefit the receptor modulation and to illustrate moieties that can commonly be used as covalent warheads. Finally, case studies are presented on designing class A, B, and C GPCR covalent allosteric modulators to demonstrate successful stories on combining allosteric modulation and covalent binding. Limitations and future perspectives are also covered.

Effects of α-Mangostin Derivatives on the Alzheimer's Disease Model of Rats and Their Mechanism: A Combination of Experimental Study and Computational Systems Pharmacology Analysis.

α-Mangostin (α-M) is a natural xanthone from the pericarp of fruit Garcinia mangostana and possesses versatile biological activities. α-M has a therapeutic potential to treat Alzheimer's disease (AD) because of its anti-inflammatory, antioxidative, and neuroprotective activities. However, the use of α-M for AD treatment is limited due to its cytotoxic activities and relatively low potency. Modifications of its chemical structure were needed to reduce its cytotoxicity and improve its therapeutic potential against AD. For this purpose, 16 α-M carbamate derivatives were synthesized. An animal model of AD was established, and the effects of AMG-1 on the spatial learning ability and memory ability were evaluated using behavioral tests. The effect on neuropathology was tested by histopathological evaluation, Nissl staining, and silver staining. Computational systems pharmacology analysis using the chemogenomics knowledgebase was applied for network studies. Compound-target, target-pathway, and target-disease networks were constructed, integrating both in silico analysis and reported experimental data. The results show that AMG-1 can demonstrate its therapeutic effects in a one-molecule, multiple-targets manner to remarkably ameliorate neurological changes and reverse behavioral deficits in AD model rats. The improved cognitive function and alleviated neuronal injury can be observed. The ability of AMG-1 to scavenge ß-amyloid in the hippocampus was validated in AD model rats.

Deep Convolutional Generative Adversarial Network (dcGAN) Models for Screening and Design of Small Molecules Targeting Cannabinoid Receptors.

A deep convolutional generative adversarial network (dcGAN) model was developed in this study to screen and design target-specific novel compounds for cannabinoid receptors. In the adversarial process of training, two models, the discriminator D and the generator G, are iteratively trained. D is trained to discover the hidden patterns among the input data to have the accurate discrimination of the authentic compounds and the "fake" compounds generated by G; G is trained to generate "fake" compounds to fool the well-trained D by optimizing the weights for matrix multiplication of data sampling. In order to determine the appropriate architecture and the input data structure for the involved convolutional neural networks (CNNs), the combinations of various network architectures and molecular fingerprints were explored. Well-developed CNN models including LeNet-5, AlexNet, ZFNet, and VGGNet were investigated. Four types of fingerprints, including MACCS, ECFP6, AtomPair, and AtomPair Count, were calculated to describe the small molecules with diverse structural characteristics. The limitation of generating fingerprints as output remains that the concrete molecular structures cannot be converted directly, while the generative models with convolutional networks provide promising opportunities to the screening of molecules and rational modifications afterward. This study demonstrated how computer-aided drug discovery could benefit from the recent advances in deep learning.

Prediction of Orthosteric and Allosteric Regulations on Cannabinoid Receptors Using Supervised Machine Learning Classifiers.

Designing highly selective compounds to protein subtypes and developing allosteric modulators targeting them are critical considerations to both drug discovery and mechanism studies for cannabinoid receptors. It is challenging but in demand to have classifiers to identify active ligands from inactive or random compounds and distinguish allosteric modulators from orthosteric ligands. In this study, supervised machine learning classifiers were built for two subtypes of cannabinoid receptors, CB1 and CB2. Three types of features, including molecular descriptors, MACCS fingerprints, and ECFP6 fingerprints, were calculated to evaluate the compound sets from diverse aspects. Deep neural networks, as well as conventional machine learning algorithms including support vector machine, naïve Bayes, logistic regression, and ensemble learning, were applied. Their performances on the classification with different types of features were compared and discussed. According to the receiver operating characteristic curves and the calculated metrics, the advantages and drawbacks of each algorithm were investigated. The feature ranking was followed to help extract useful knowledge about critical molecular properties, substructural keys, and circular fingerprints. The extracted features will then facilitate the research on cannabinoid receptors by providing guidance on preferred properties for compound modification and novel scaffold design. Besides using conventional molecular docking studies for compound virtual screening, machine-learning-based decision-making models provide alternative options. This study can be of value to the application of machine learning in the area of drug discovery and compound development.

Significantly different effects of tetrahydroberberrubine enantiomers on dopamine D1/D2 receptors revealed by experimental study and integrated in silico simulation.

Tetrahydroberberrubine (TU), an active tetrahydroprotoberberines (THPBs), is gaining increasing popularity as a potential candidate for treatment of anxiety and depression. One of its two enantiomers, l-TU, has been reported to be an antagonist of both D1 and D2 receptors, but the functional activity of the other enantiomer, d-TU, is still unknown. In this study, experiments were combined with in silico molecular simulations to (1) confirm and discover the functional activities of l-TU and d-TU, and (2) systematically evaluate the molecular mechanisms beyond the experimental observations. l-TU proved to be an antagonist of both D1 and D2 receptors (IC50 = 385 nM and 985 nM, respectively), while d-TU shows no affinity against either D1 or D2 receptor, based on the cAMP assay (D1 receptor) and calcium flux assay (D2 receptor). Results from both flexible-ligand docking studies and molecular dynamic (MD) simulations provided insights at the atomic level. The l-TU-bound structures predicted by MD (1) undergo an outward rotation of the extracellular helical bundles; (2) have an enlarged orthosteric binding pocket; and (3) have a central toggle switch that is prevented from rotating freely. These features are unique to the l-TU enantiomer and provide an explanation for its antagonistic behavior toward both D1 and D2 receptors. The present study provides new sight on the structural and functional relationships of l-TU and d-TU binding to dopamine receptors, and provides guidance to the rational design of novel molecules targeting these two dopamine receptors in the future.

Computational systems pharmacology analysis of cannabidiol: a combination of chemogenomics-knowledgebase network analysis and integrated in silico modeling and simulation.

With treatment benefits in both the central nervous system and the peripheral system, the medical use of cannabidiol (CBD) has gained increasing popularity. Given that the therapeutic mechanisms of CBD are still vague, the systematic identification of its potential targets, signaling pathways, and their associations with corresponding diseases is of great interest for researchers. In the present work, chemogenomics-knowledgebase systems pharmacology analysis was applied for systematic network studies to generate CBD-target, target-pathway, and target-disease networks by combining both the results from the in silico analysis and the reported experimental validations. Based on the network analysis, three human neuro-related rhodopsin-like GPCRs, i.e., 5-hydroxytryptamine receptor 1 A (5HT1A), delta-type opioid receptor (OPRD) and G protein-coupled receptor 55 (GPR55), were selected for close evaluation. Integrated computational methodologies, including homology modeling, molecular docking, and molecular dynamics simulation, were used to evaluate the protein-CBD binding modes. A CBD-preferred pocket consisting of a hydrophobic cavity and backbone hinges was proposed and tested for CBD-class A GPCR binding. Finally, the neurophysiological effects of CBD were illustrated at the molecular level, and dopamine receptor 3 (DRD3) was further predicted to be an active target for CBD.

Correction to: Deep Learning for Drug Design: an Artificial Intelligence Paradigm for Drug Discovery in the Big Data Era.

The name of the corresponding author should be 'Xiang-Qun Xie', rather than 'Xiang-Qun Sean Xie'.

Computational Fragment-Based Drug Design: Current Trends, Strategies, and Applications.

Fragment-based drug design (FBDD) has become an effective methodology for drug development for decades. Successful applications of this strategy brought both opportunities and challenges to the field of Pharmaceutical Science. Recent progress in the computational fragment-based drug design provide an additional approach for future research in a time- and labor-efficient manner. Combining multiple in silico methodologies, computational FBDD possesses flexibilities on fragment library selection, protein model generation, and fragments/compounds docking mode prediction. These characteristics provide computational FBDD superiority in designing novel and potential compounds for a certain target. The purpose of this review is to discuss the latest advances, ranging from commonly used strategies to novel concepts and technologies in computational fragment-based drug design. Particularly, in this review, specifications and advantages are compared between experimental and computational FBDD, and additionally, limitations and future prospective are discussed and emphasized.

Deep Learning for Drug Design: an Artificial Intelligence Paradigm for Drug Discovery in the Big Data Era.

Over the last decade, deep learning (DL) methods have been extremely successful and widely used to develop artificial intelligence (AI) in almost every domain, especially after it achieved its proud record on computational Go. Compared to traditional machine learning (ML) algorithms, DL methods still have a long way to go to achieve recognition in small molecular drug discovery and development. And there is still lots of work to do for the popularization and application of DL for research purpose, e.g., for small molecule drug research and development. In this review, we mainly discussed several most powerful and mainstream architectures, including the convolutional neural network (CNN), recurrent neural network (RNN), and deep auto-encoder networks (DAENs), for supervised learning and nonsupervised learning; summarized most of the representative applications in small molecule drug design; and briefly introduced how DL methods were used in those applications. The discussion for the pros and cons of DL methods as well as the main challenges we need to tackle were also emphasized.

Integrated In Silico Fragment-Based Drug Design: Case Study with Allosteric Modulators on Metabotropic Glutamate Receptor 5.

GPCR allosteric modulators target at the allosteric binding pockets of G protein-coupled receptors (GPCRs) with indirect influence on the effects of an orthosteric ligand. Such modulators exhibit significant advantages compared to the corresponding orthosteric ligands, including better chemical tractability or physicochemical properties, improved selectivity, and reduced risk of oversensitization towards their receptors. Metabotropic glutamate receptor 5 (mGlu5), a member of class C GPCRs, is a promising therapeutic target for treating many central nervous system diseases. The crystal structure of mGlu5 in the complex with the negative allosteric modulator mavoglurant was recently reported, providing a fundamental model for designing new allosteric modulators. Computational fragment-based drug discovery represents a powerful scaffold-hopping and lead structure-optimization tool for drug design. In the present work, a set of integrated computational methodologies was first used, such as fragment library generation and retrosynthetic combinatorial analysis procedure (RECAP) for novel compound generation. Then, the compounds generated were assessed by benchmark dataset verification, docking studies, and QSAR model simulation. Subsequently, structurally diverse compounds, with reported or unreported scaffolds, can be observed from top 20 in silico synthesized compounds, which were predicted to be potential mGlu5 modulators. In silico compounds with reported scaffolds may fill SAR holes in known, patented series of mGlu5 modulators. And the generation of compounds without reported tests on mGluR indicates that our approach is doable for exploring and designing novel compounds. Our case study of designing allosteric modulators on mGlu5 demonstrated that the established computational fragment-based approach is a useful methodology for facilitating new compound design in the future.