MELLODDY: Cross-pharma Federated Learning at Unprecedented Scale Unlocks Benefits in QSAR without Compromising Proprietary Information.

Federated multipartner machine learning has been touted as an appealing and efficient method to increase the effective training data volume and thereby the predictivity of models, particularly when the generation of training data is resource-intensive. In the landmark MELLODDY project, indeed, each of ten pharmaceutical companies realized aggregated improvements on its own classification or regression models through federated learning. To this end, they leveraged a novel implementation extending multitask learning across partners, on a platform audited for privacy and security. The experiments involved an unprecedented cross-pharma data set of 2.6+ billion confidential experimental activity data points, documenting 21+ million physical small molecules and 40+ thousand assays in on-target and secondary pharmacodynamics and pharmacokinetics. Appropriate complementary metrics were developed to evaluate the predictive performance in the federated setting. In addition to predictive performance increases in labeled space, the results point toward an extended applicability domain in federated learning. Increases in collective training data volume, including by means of auxiliary data resulting from single concentration high-throughput and imaging assays, continued to boost predictive performance, albeit with a saturating return. Markedly higher improvements were observed for the pharmacokinetics and safety panel assay-based task subsets.

Deep generative models for ligand-based de novo design applied to multi-parametric optimization.

Multi-parameter optimization (MPO) is a major challenge in new chemical entity (NCE) drug discovery. Recently, promising results were reported for deep learning generative models applied to de novo molecular design, but, to our knowledge, until now no report was made of the value of this new technology for addressing MPO in an actual drug discovery project. In this study, we demonstrate the benefit of applying AI technology in a real drug discovery project. We evaluate the potential of a ligand-based de novo design technology using deep learning generative models to accelerate the obtention of lead compounds meeting 11 different biological activity objectives simultaneously. Using the initial dataset of the project, we built QSAR models for all the 11 objectives, with moderate to high performance (precision between 0.67 and 1.0 on an independent test set). Our DL-based AI de novo design algorithm, combined with the QSAR models, generated 150 virtual compounds predicted as active on all objectives. Eleven were synthetized and tested. The AI-designed compounds met 9.5 objectives on average (i.e., 86% success rate) versus 6.4 (i.e., 58% success rate) for the initial molecules measured on all objectives. One of the AI-designed molecules was active on all 11 measured objectives, and two were active on 10 objectives while being in the error margin of the assay for the last one. The AI algorithm designed compounds with functional groups, which, although being rare or absent in the initial dataset, turned out to be highly beneficial for the MPO.

Development of a chemogenomics library for phenotypic screening.

With the development of advanced technologies in cell-based phenotypic screening, phenotypic drug discovery (PDD) strategies have re-emerged as promising approaches in the identification and development of novel and safe drugs. However, phenotypic screening does not rely on knowledge of specific drug targets and needs to be combined with chemical biology approaches to identify therapeutic targets and mechanisms of actions induced by drugs and associated with an observable phenotype. In this study, we developed a system pharmacology network integrating drug-target-pathway-disease relationships as well as morphological profile from an existing high content imaging-based high-throughput phenotypic profiling assay known as "Cell Painting". Furthermore, from this network, a chemogenomic library of 5000 small molecules that represent a large and diverse panel of drug targets involved in diverse biological effects and diseases has been developed. Such a platform and a chemogenomic library could assist in the target identification and mechanism deconvolution of some phenotypic assays. The usefulness of the platform is illustrated through examples.

Shape-Based Descriptors for Efficient Structure-Based Fragment Growing.

Structure-based fragment growing is one of the key techniques in fragment-based drug design. Fragment growing is commonly practiced based on structural and biophysical data. Computational workflows are employed to predict which fragment elaborations could lead to high-affinity binders. Several such workflows exist but many are designed to be long running noninteractive systems. Shape-based descriptors have been proven to be fast and perform well at virtual-screening tasks. They could, therefore, be applied to the fragment-growing problem to enable an interactive fragment-growing workflow. In this work, we describe and analyze the use of specific shape-based directional descriptors for the task of fragment growing. The performance of these descriptors that we call ray volume matrices (RVMs) is evaluated on two data sets containing protein-ligand complexes. While the first set focuses on self-growing, the second measures practical performance in a cross-growing scenario. The runtime of screenings using RVMs as well as their robustness to three dimensional perturbations is also investigated. Overall, it can be shown that RVMs are useful to prefilter fragment candidates. For up to 84% of the 3299 generated self-growing cases and for up to 66% of the 326 generated cross-growing cases, RVMs could create poses with less than 2 Å root-mean-square deviation to the crystal structure with average query speeds of around 30,000 conformations per second. This opens the door for fast explorative screenings of fragment libraries.

Importance of the second extracellular loop for melatonin MT1 receptor function and absence of melatonin binding in GPR50.

BACKGROUND AND PURPOSE: Recent crystal structures of GPCRs have emphasized the previously unappreciated role of the second extracellular (E2) loop in ligand binding and gating and receptor activation. Here, we have assessed the role of the E2 loop in the activation of the melatonin MT1 receptor and in the inactivation of the closely related orphan receptor GPR50. EXPERIMENTAL APPROACH: Chimeric MT1 -GPR50 receptors were generated and functionally analysed in terms of 2-[125 I]iodomelatonin binding, Gi /cAMP signalling and ß-arrestin2 recruitment. We also used computational molecular dynamics (MD) simulations. KEY RESULTS: MD simulations of 300 ns revealed (i) the tight hairpin structure of the E2 loop of the MT1 receptor (ii) the most suitable features for melatonin binding in MT1 receptors and (iii) major predicted rearrangements upon MT1 receptor activation, stabilizing interaction networks between Phe179 or Gln181 in the E2 loop and transmembrane helixes 5 and 6. Functional assays confirmed these predictions, because reciprocal replacement of MT1 and GPR50 residues/domains led to the predicted loss- and gain-of-melatonin action of MT1 receptors and GPR50 respectively. CONCLUSIONS AND IMPLICATIONS: Our work demonstrated the crucial role of the E2 loop for MT1 receptor and GPR50 function by proposing a model in which the E2 loop is important in stabilizing active MT1 receptor conformations and by showing how evolutionary processes appear to have selected for modifications in the E2 loop in order to make GPR50 unresponsive to melatonin. LINKED ARTICLES: This article is part of a themed section on Recent Developments in Research of Melatonin and its Potential Therapeutic Applications. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v175.16/issuetoc.

Design and synthesis of novel N-sulfonyl-2-indoles that behave as 5-HT6 receptor ligands with significant selectivity for D3 over D2 receptors.

All clinically-used antipsychotics display similar affinity for both D2 (D2R) and D3 (D3R) receptors, and they likewise act as 5-HT2A receptor antagonists. They provide therapeutic benefit for positive symptoms, but no marked or consistent improvement in neurocognitive, social cognitive or negative symptoms. Since blockade of D3 and 5-HT6 (5-HT6R) receptors enhances neurocognition and social cognition, and potentially improves negative symptoms, a promising approach for improved treatment for schizophrenia would be to develop drugs that preferentially act at D3R versus D2R and likewise recognize 5-HT6R. Starting from the high affinity 5-HT6R ligands I and II, we identified compounds 11a and 14b that behave as 5-HT6R ligands with significant selectivity for D3R over D2R.

Synthesis and pharmacological evaluation of thieno[2,3-b]pyridine derivatives as novel c-Src inhibitors.

Among the recently investigated targets for cancer therapy is the c-Src non-receptor tyrosine kinase. Indeed research around deregulated activity of this enzyme has proven its role in tumor progression, while the beneficial effects of c-Src inhibitors in several pathological models has also been demonstrated. We report here the preparation and pharmacological profile of a novel series of c-Src inhibitors that was elaborated around a 3-amino-thieno[2,3-b]pyridine discovered during an HTS campaign. c-Src enzyme inhibition and c-Src inhibition were investigated in a series of related compounds derived from the initial hit. Molecular modeling as well as X-ray studies on one active compound allowed us to hypothesize on ligand orientation and interactions within the ATP hydrophobic pocket. Design and synthesis of structural analogs then led to new ligands possessing quite efficient enzymatic and c-Src inhibition. The structure-activity elements disclosed in this study shed light on the role played by substituents on the thienopyridine ring as well as the impact of other aromatic moieties in the molecule when interacting with the enzyme.

Novel structural insights for drug design of selective 5-HT(2C) inverse agonists from a ligand-biased receptor model.

Structure-based design of compounds targeting monoamine receptors, within the class-A G-protein coupled receptors, has been enriched by the recent crystallization of the ß1 and ß2 adrenoceptors. On the basis of ligand-biased homology modeling and docking-scoring calculations, a ritanserin-biased 5-HT(2C) receptor model has been built and used in a highly efficient virtual screening protocol to discriminate specifically 5-HT(2C) inverse agonists in a fuzzy dataset including hundreds of compounds with known experimental values of 5-HT(2C) affinity and activity. The resulting fingerprint of interaction displays hotspots in the third transmembrane α-helix and the second extracellular loop selectively bound by most 5-HT(2C) inverse agonists.

Pyridyldithiafulvenes as precursors of coordination-driven self-assembled redox active macrocycle.

A novel redox active macrocycle including two vinylogous tetrathiafulvalenes (TTFVs) and two molybdenum tetracarbonyl fragments has been synthezised thanks to the coordination-driven self-assembly of complementary angular derivatives. Pyridyl vinylogous TTFVs have been deliberately elaborated for that purpose, using the oxidative coupling of pyridyldithiafulvenes (DTF). Cyclic voltammetry, IR and NMR spectroscopies, and single-crystal X-ray crystallography of the target molecules have been investigated.

Studies of the melatonin binding site location onto quinone reductase 2 by directed mutagenesis.

Melatonin is a neurohormone implicated in both biorhythm synchronization and neuroprotection from oxidative stress. Its functions are mediated by two G-protein-coupled-receptors (MT1 and MT2) and MT3, which corresponds to quinone oxidoreductase 2 (QR2). To determine the binding site of QR2 for melatonin, point mutations of residues crucial for the enzymatic activity of hQR2 were performed. The substitution of the hydrophobic residues Phe126, Ile128 and Phe178 by tyrosines at the active site significantly increased enzymatic activity and decreased the affinity of a structural analog of melatonin, the 2[(125)I]iodo-MCANAT. The mutation of residues implicated in zinc chelating (His(173); His(177)) had no effect on radioligand binding. Destabilisation of the cofactor FAD by mutation N18E showed that 2[(125)I]iodo-MCANAT binding was closely linked to the conformational integrity of human QR2. Surprisingly, the mutations C222F and N161A, which are distant from the determined binding site of the ligand, increased the affinity of 2[(125)I]iodo-MCANAT for hQR2. What seems to better explain the binding variations among the mutants are the activity recorded with BNAH and coenzyme Q1. Various hypotheses are discussed based on the various parameters used in the study: nature of the substrates and co-substrates and nature of the amino acid changes. This study, which constitutes the first structural analysis of hQR2, should enable to better understand the biological role of melatonin on this enzyme and particularly, the discrepancies between the pharmacologies of the melatonin binding site (MT3) and the QR2 catalytic activity.