Targeting non-alcoholic fatty liver disease: Design, X-ray co-crystal structure and synthesis of 'first-in-kind' inhibitors of serine/threonine kinase25.

We describe the synthesis of a series of 3-t-butyl 5-aminopyrazole p-substituted arylamides as inhibitors of serine-threonine25 (STK25), an enzyme implicated in the progression of non-alcoholic fatty liver disease (NAFLD). Appending a p-N-pyrrolidinosulphonamide group to the arylamide group led to a 'first-in kind' inhibitor with IC50 = 228 nM. A co-crystal structure with STK 25 revealed productive interactions which were also reproduced using molecular docking. A new series of triazolo dihydro oxazine carboxamides of 3-t-butyl 5-aminopyrazole was not active against STK25.

Artificial intelligence-enhanced drug design and development: Toward a computational precision medicine.

Artificial Intelligence (AI) relies upon a convergence of technologies with further synergies with life science technologies to capture the value of massive multi-modal data in the form of predictive models supporting decision-making. AI and machine learning (ML) enhance drug design and development by improving our understanding of disease heterogeneity, identifying dysregulated molecular pathways and therapeutic targets, designing and optimizing drug candidates, as well as evaluating in silico clinical efficacy. By providing an unprecedented level of knowledge on both patient specificities and drug candidate properties, AI is fostering the emergence of a computational precision medicine allowing the design of therapies or preventive measures tailored to the singularities of individual patients in terms of their physiology, disease features, and exposure to environmental risks.

Improved Lipophilicity and Aqueous Solubility Prediction with Composite Graph Neural Networks.

The accurate prediction of molecular properties, such as lipophilicity and aqueous solubility, are of great importance and pose challenges in several stages of the drug discovery pipeline. Machine learning methods, such as graph-based neural networks (GNNs), have shown exceptionally good performance in predicting these properties. In this work, we introduce a novel GNN architecture, called directed edge graph isomorphism network (D-GIN). It is composed of two distinct sub-architectures (D-MPNN, GIN) and achieves an improvement in accuracy over its sub-architectures employing various learning, and featurization strategies. We argue that combining models with different key aspects help make graph neural networks deeper and simultaneously increase their predictive power. Furthermore, we address current limitations in assessment of deep-learning models, namely, comparison of single training run performance metrics, and offer a more robust solution.

New machine learning and physics-based scoring functions for drug discovery.

Scoring functions are essential for modern in silico drug discovery. However, the accurate prediction of binding affinity by scoring functions remains a challenging task. The performance of scoring functions is very heterogeneous across different target classes. Scoring functions based on precise physics-based descriptors better representing protein-ligand recognition process are strongly needed. We developed a set of new empirical scoring functions, named DockTScore, by explicitly accounting for physics-based terms combined with machine learning. Target-specific scoring functions were developed for two important drug targets, proteases and protein-protein interactions, representing an original class of molecules for drug discovery. Multiple linear regression (MLR), support vector machine and random forest algorithms were employed to derive general and target-specific scoring functions involving optimized MMFF94S force-field terms, solvation and lipophilic interactions terms, and an improved term accounting for ligand torsional entropy contribution to ligand binding. DockTScore scoring functions demonstrated to be competitive with the current best-evaluated scoring functions in terms of binding energy prediction and ranking on four DUD-E datasets and will be useful for in silico drug design for diverse proteins as well as for specific targets such as proteases and protein-protein interactions. Currently, the MLR DockTScore is available at www.dockthor.lncc.br .

A compact review of molecular property prediction with graph neural networks.

As graph neural networks are becoming more and more powerful and useful in the field of drug discovery, many pharmaceutical companies are getting interested in utilizing these methods for their own in-house frameworks. This is especially compelling for tasks such as the prediction of molecular properties which is often one of the most crucial tasks in computer-aided drug discovery workflows. The immense hype surrounding these kinds of algorithms has led to the development of many different types of promising architectures and in this review we try to structure this highly dynamic field of AI-research by collecting and classifying 80 GNNs that have been used to predict more than 20 molecular properties using 48 different datasets.

Cheminformatics approach to exploring and modeling trait-associated metabolite profiles.

Developing predictive and transparent approaches to the analysis of metabolite profiles across patient cohorts is of critical importance for understanding the events that trigger or modulate traits of interest (e.g., disease progression, drug metabolism, chemical risk assessment). However, metabolites' chemical structures are still rarely used in the statistical modeling workflows that establish these trait-metabolite relationships. Herein, we present a novel cheminformatics-based approach capable of identifying predictive, interpretable, and reproducible trait-metabolite relationships. As a proof-of-concept, we utilize a previously published case study consisting of metabolite profiles from non-small-cell lung cancer (NSCLC) adenocarcinoma patients and healthy controls. By characterizing each structurally annotated metabolite using both computed molecular descriptors and patient metabolite concentration profiles, we show that these complementary features enhance the identification and understanding of key metabolites associated with cancer. Ultimately, we built multi-metabolite classification models for assessing patients' cancer status using specific groups of metabolites identified based on high structural similarity through chemical clustering. We subsequently performed a metabolic pathway enrichment analysis to identify potential mechanistic relationships between metabolites and NSCLC adenocarcinoma. This cheminformatics-inspired approach relies on the metabolites' structural features and chemical properties to provide critical information about metabolite-trait associations. This method could ultimately facilitate biological understanding and advance research based on metabolomics data, especially with respect to the identification of novel biomarkers.

Identifying individual risk rare variants using protein structure guided local tests (POINT).

Rare variants are of increasing interest to genetic association studies because of their etiological contributions to human complex diseases. Due to the rarity of the mutant events, rare variants are routinely analyzed on an aggregate level. While aggregation analyses improve the detection of global-level signal, they are not able to pinpoint causal variants within a variant set. To perform inference on a localized level, additional information, e.g., biological annotation, is often needed to boost the information content of a rare variant. Following the observation that important variants are likely to cluster together on functional domains, we propose a protein structure guided local test (POINT) to provide variant-specific association information using structure-guided aggregation of signal. Constructed under a kernel machine framework, POINT performs local association testing by borrowing information from neighboring variants in the 3-dimensional protein space in a data-adaptive fashion. Besides merely providing a list of promising variants, POINT assigns each variant a p-value to permit variant ranking and prioritization. We assess the selection performance of POINT using simulations and illustrate how it can be used to prioritize individual rare variants in PCSK9, ANGPTL4 and CETP in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) clinical trial data.

Confirmation of high-throughput screening data and novel mechanistic insights into VDR-xenobiotic interactions by orthogonal assays.

High throughput screening (HTS) programs have demonstrated that the Vitamin D receptor (VDR) is activated and/or antagonized by a wide range of structurally diverse chemicals. In this study, we examined the Tox21 qHTS data set generated against VDR for reproducibility and concordance and elucidated functional insights into VDR-xenobiotic interactions. Twenty-one potential VDR agonists and 19 VDR antagonists were identified from a subset of >400 compounds with putative VDR activity and examined for VDR functionality utilizing select orthogonal assays. Transient transactivation assay (TT) using a human VDR plasmid and Cyp24 luciferase reporter construct revealed 20/21 active VDR agonists and 18/19 active VDR antagonists. Mammalian-2-hybrid assay (M2H) was then used to evaluate VDR interactions with co-activators and co-regulators. With the exception of a select few compounds, VDR agonists exhibited significant recruitment of co-regulators and co-activators whereas antagonists exhibited considerable attenuation of recruitment by VDR. A unique set of compounds exhibiting synergistic activity in antagonist mode and no activity in agonist mode was identified. Cheminformatics modeling of VDR-ligand interactions were conducted and revealed selective ligand VDR interaction. Overall, data emphasizes the molecular complexity of ligand-mediated interactions with VDR and suggest that VDR transactivation may be a target site of action for diverse xenobiotics.

In silico Predicted Glucose-1-phosphate Uridylyltransferase (GalU) Inhibitors Block a Key Pathway Required for Listeria Virulence.

Peptidoglycan walls of gram positive bacteria are functionalized by glycopolymers called wall teichoic acid (WTA). In Listeria monocytogenes, multiple enzymes including the glucose-1-phosphate uridylyltransferase (GalU) were identified as mandatory for WTA galactosylation, so that the inhibition of GalU is associated with a significant attenuation of Listeria virulence. Herein, we report on a series of in silico predicted GalU inhibitors identified using structure-based virtual screening and experimentally validated to be effective in blocking the WTA galactosylation pathway in vitro. Several hits such as C04, a pyrimidinyl benzamide, afforded promising experimental potencies. This proof-of-concept study opens new perspectives for the development of potent and selective GalU inhibitors of high interest to attenuate Listeria virulence. It also underscores the high relevance of using molecular modeling for facilitating the identification of bacterial virulence attenuators and more generally antibacterials.

Cheminformatics Analysis of Dynamic WNK-Inhibitor Interactions.

The With-No-Lysine (WNK) serine/threonine kinase family constitutes a unique and distinctive branch of the kinome. The four proteins of this family (WNK1/2/3/4) are involved in blood pressure regulation, body fluid, and electrolyte homeostasis. Herein, we modeled and analyzed the binding modes of all publicly-available small orthosteric and allosteric binders (including WNK463 and WNK467) experimentally tested towards any of the WNK family member. To do so, we relied on state-of-the-art cheminformatics approaches including structure-based molecular docking and molecular dynamics simulations. In particular, we computed and analyzed the (i) molecular selectivity of known inhibitors when docked in the binding site of each WNK family member, (ii) the dynamic WNK-inhibitor interactions at both orthosteric and allosteric sites to derive new structure-activity relationships, and (iii) the key specific interactions present in each binding site. This study reports on the first, cheminformatics-powered analysis of the entire chemical space of known WNK inhibitors. We discuss the conservation of critical WNK-inhibitor interactions and the existence of isoform-specific interactions that could enable the rational design of more potent and selective WNK binders.

Weaver's historic accessible collection of synthetic dyes: a cheminformatics analysis.

We present the Max Weaver Dye Library, a collection of ∼98 000 vials of custom-made and largely sparingly water-soluble dyes. Two years ago, the Eastman Chemical Company donated the library to North Carolina State University. This unique collection of chemicals, housed in the College of Textiles, also includes tens of thousands of fabric samples dyed using some of the library's compounds. Although the collection lies at the core of hundreds of patented inventions, the overwhelming majority of this chemical treasure trove has never been published or shared outside of a small group of scientists. Thus, the goal of this donation was to make this chemical collection, and associated data, available to interested parties in the research community. To date, we have digitized a subset of 2700 dyes which allowed us to start the constitutional and structural analysis of the collection using cheminformatics approaches. Herein, we open the discussion regarding the research opportunities offered by this unique library.

Privileged Substructures to Modulate Protein-Protein Interactions.

Given the difficulties to identify chemical probes that can modulate protein-protein interactions (PPIs), actors in the field have started to agree on the necessity to use PPI-tailored screening chemical collections. However, which type of scaffolds may promote the binding of compounds to PPI targets remains unclear. In this big data analysis, we have identified a list of privileged chemical substructures that are most often observed within inhibitors of PPIs. Using molecular frameworks as a way to perceive chemical substructures with the combination of an experimental and a machine-learning based predicted data set of iPPI compounds, we propose a list of privileged substructures in the form of scaffolds and chemical moieties that can be substantially chemically functionalized and do not present any toxicophore nor pan-assay interference (PAINS) alerts. We think that such chemical guidance will be valuable for medicinal chemists in their attempt to identify initial quality chemical probes on PPI targets.

Cheminformatics Modeling of Amine Solutions for Assessing their CO2 Absorption Properties.

As stricter regulations on CO2 emissions are adopted worldwide, identifying efficient chemical processes to capture and recycle CO2 is of critical importance for industry. The most common process known as amine scrubbing suffers from the lack of available amine solutions capable of capturing CO2 efficiently. Tertiary amines characterized by low heats of reaction are considered good candidates but their absorption properties can significantly differ from one analogue to another despite high structural similarity. Herein, after collecting and curating experimental data from the literature, we have built a modeling set of 41 amine structures with their absorption properties. Then we analyzed their chemical composition using molecular descriptors and non-supervised clustering. Furthermore, we developed a series of quantitative structure-property relationships (QSPR) to assess amines' CO2 absorption properties from their structural characteristics. These models afforded reasonable prediction performances (e. g., Q2LOO =0.63 for CO2 absorption amount) even though they are solely based on 2D chemical descriptors and individual machine learning techniques (random forest and neural network). Overall, we believe the chemical analysis and the series of QSPR models presented in this proof-of-concept study represent new knowledge and innovative tools that could be very useful for screening and prioritizing hypothetical amines to be synthesized and tested experimentally for their CO2 absorption properties.

Imbalance in chemical space: How to facilitate the identification of protein-protein interaction inhibitors.

Protein-protein interactions (PPIs) play vital roles in life and provide new opportunities for therapeutic interventions. In this large data analysis, 3,300 inhibitors of PPIs (iPPIs) were compared to 17 reference datasets of collectively ~566,000 compounds (including natural compounds, existing drugs, active compounds on conventional targets, etc.) using a chemoinformatics approach. Using this procedure, we showed that comparable classes of PPI targets can be formed using either the similarity of their ligands or the shared properties of their binding cavities, constituting a proof-of-concept that not only can binding pockets be used to group PPI targets, but that these pockets certainly condition the properties of their corresponding ligands. These results demonstrate that matching regions in both chemical space and target space can be found. Such identified classes of targets could lead to the design of PPI-class-specific chemical libraries and therefore facilitate the development of iPPIs to the stage of drug candidates.

iPPI-DB: an online database of modulators of protein-protein interactions.

In order to boost the identification of low-molecular-weight drugs on protein-protein interactions (PPI), it is essential to properly collect and annotate experimental data about successful examples. This provides the scientific community with the necessary information to derive trends about privileged physicochemical properties and chemotypes that maximize the likelihood of promoting a given chemical probe to the most advanced stages of development. To this end we have developed iPPI-DB (freely accessible at http://www.ippidb.cdithem.fr), a database that contains the structure, some physicochemical characteristics, the pharmacological data and the profile of the PPI targets of several hundreds modulators of protein-protein interactions. iPPI-DB is accessible through a web application and can be queried according to two general approaches: using physicochemical/pharmacological criteria; or by chemical similarity to a user-defined structure input. In both cases the results are displayed as a sortable and exportable datasheet with links to external databases such as Uniprot, PubMed. Furthermore each compound in the table has a link to an individual ID card that contains its physicochemical and pharmacological profile derived from iPPI-DB data. This includes information about its binding data, ligand and lipophilic efficiencies, location in the PPI chemical space, and importantly similarity with known drugs, and links to external databases like PubChem, and ChEMBL.

Stabilization of protein-protein interaction complexes through small molecules.

Most of the small molecules that have been identified thus far to modulate protein-protein interactions (PPIs) are inhibitors. Another promising way to interfere with PPI-associated biological processes is to promote PPI stabilization. Even though PPI stabilizers are still scarce, stabilization of PPIs by small molecules is gaining momentum and offers new pharmacological options. Therefore, we have performed a literature survey of PPI stabilization using small molecules. From this, we propose a classification of PPI stabilizers based on their binding mode and the architecture of the complex to facilitate the structure-based design of stabilizers.

In silico design of low molecular weight protein-protein interaction inhibitors: Overall concept and recent advances.

Protein-protein interactions (PPIs) are carrying out diverse functions in living systems and are playing a major role in the health and disease states. Low molecular weight (LMW) "drug-like" inhibitors of PPIs would be very valuable not only to enhance our understanding over physiological processes but also for drug discovery endeavors. However, PPIs were deemed intractable by LMW chemicals during many years. But today, with the new experimental and in silico technologies that have been developed, about 50 PPIs have already been inhibited by LMW molecules. Here, we first focus on general concepts about protein-protein interactions, present a consensual view about ligandable pockets at the protein interfaces and the possibilities of using fast and cost effective structure-based virtual screening methods to identify PPI hits. We then discuss the design of compound collections dedicated to PPIs. Recent financial analyses of the field suggest that LMW PPI modulators could be gaining momentum over biologics in the coming years supporting further research in this area.

Which three-dimensional characteristics make efficient inhibitors of protein-protein interactions?

The specific properties of protein-protein interactions (PPI) (flat, large and hydrophobic) make them harder to tackle with low-molecular-weight compounds. Learning from the properties of successful examples of PPI interface inhibitors (iPPI) at earlier stages of developments, has been pinpointed as a powerful strategy to circumvent this trend. To this end, we have computationally analyzed the bioactive conformations of iPPI and those of inhibitors of conventional targets (e.g enzymes) to highlight putative iPPI 3D characteristics. Most noticeably, the essential property revealed by this study illustrates how efficiently iPPI manages to bind to the hydrophobic patch often present at the core of protein interfaces. The newly identified properties were further confirmed as characteristics of iPPI using much larger data sets (e.g iPPI-DB, www.ippidb.cdithem.fr ). Interestingly, the absence of correlation of such properties with the hydrophobicity and the size of the compounds opens new ways to design potent iPPI with better pharmacokinetic features.

Drug-Like Protein-Protein Interaction Modulators: Challenges and Opportunities for Drug Discovery and Chemical Biology.

[Formula: see text] Fundamental processes in living cells are largely controlled by macromolecular interactions and among them, protein-protein interactions (PPIs) have a critical role while their dysregulations can contribute to the pathogenesis of numerous diseases. Although PPIs were considered as attractive pharmaceutical targets already some years ago, they have been thus far largely unexploited for therapeutic interventions with low molecular weight compounds. Several limiting factors, from technological hurdles to conceptual barriers, are known, which, taken together, explain why research in this area has been relatively slow. However, this last decade, the scientific community has challenged the dogma and became more enthusiastic about the modulation of PPIs with small drug-like molecules. In fact, several success stories were reported both, at the preclinical and clinical stages. In this review article, written for the 2014 International Summer School in Chemoinformatics (Strasbourg, France), we discuss in silico tools (essentially post 2012) and databases that can assist the design of low molecular weight PPI modulators (these tools can be found at www.vls3d.com). We first introduce the field of protein-protein interaction research, discuss key challenges and comment recently reported in silico packages, protocols and databases dedicated to PPIs. Then, we illustrate how in silico methods can be used and combined with experimental work to identify PPI modulators.

iPPI-DB: a manually curated and interactive database of small non-peptide inhibitors of protein-protein interactions.

The development of small molecule drugs targeting protein-protein interactions (PPI) represents a major challenge, in part owing to the misunderstanding of the PPI chemical space. To this end, we have manually collected the structures, the physicochemical and pharmacological profiles of 1650 PPI inhibitors across 13 families of PPI targets in a database named iPPI-DB. To access iPPI-DB, we propose a user-friendly web application (www.ippidb.cdithem.fr) with customizable queries and intuitive visualizing functionalities for associated properties of the compounds. This could assist scientists to design the next generation of PPI drugs. In this review, we describe iPPI-DB in the context of other low molecular weight molecule databases.