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.

Potentiation of Muscarinic M3 Receptor Activation through a New Allosteric Site with a Novel Positive Allosteric Modulator ASP8302.

Muscarinic M3 (M3) receptors mediate a wide range of acetylcholine (ACh)-induced functions, including visceral smooth-muscle contraction and glandular secretion. Positive allosteric modulators (PAMs) can avoid various side effects of muscarinic agonists with their spatiotemporal receptor activation control and potentially better subtype selectivity. However, the mechanism of allosteric modulation of M3 receptors is not fully understood, presumably because of the lack of a potent and selective PAM. In this study, we investigated the pharmacological profile of ASP8302, a novel PAM of M3 receptors, and explored the principal site of amino-acid sequences in the human M3 receptor required for the potentiation of receptor activation. In cells expressing human M3 and M5 receptors, ASP8302 shifted the concentration-response curve (CRC) for carbachol to the lower concentrations with no significant effects on other subtypes. In a binding study with M3 receptor-expressing membrane, ASP8302 also shifted the CRC for ACh without affecting the binding of orthosteric agonists. Similar shifts in the CRC of contractions by multiple stimulants were also confirmed in isolated human bladder strips. Mutagenesis analysis indicated no interaction between ASP8302 and previously reported allosteric sites; however, it identified threonine 230 as the amino acid essential for the PAM effect of ASP8302. These results demonstrate that ASP8302 enhances the activation of human M3 receptors by interacting with a single amino acid distinct from the reported allosteric sites. Our findings suggest not only a novel allosteric site of M3 receptors but also the potential application of ASP8302 to diseases caused by insufficient M3 receptor activation. SIGNIFICANCE STATEMENT: The significance of this study is that the novel M3 receptor positive allosteric modulator ASP8302 enhances the activation of human M3 receptor by interacting with a residue distinct from the reported allosteric sites. The finding of Thr230 as a novel amino acid involved in the allosteric modulation of M3 receptors provides significant insight into further research of the mechanism of allosteric modulation of M3 and other muscarinic receptors.

Design, Synthesis, and Structure-Activity Relationships Study of N-Pyrimidyl/Pyridyl-2-thiazolamine Analogues as Novel Positive Allosteric Modulators of M3 Muscarinic Acetylcholine Receptor.

The M3 muscarinic acetylcholine receptor (mAChR) plays an essential pharmacological role in mediating a broad range of actions of acetylcholine (ACh) released throughout the periphery and central nerve system (CNS). Nevertheless, its agonistic functions remain unclear due to the lack of available subtype-selective agonists or positive allosteric modulators (PAMs). In the course of our extended structure-activity relationships (SARs) study on 2-acylaminothiazole derivative 1, a previously reported PAM of the M3 mAChR, we successfully identified N-pyrimidyl/pyridyl-2-thiazolamine analogues as new scaffolds. The SARs study was rationalized using conformational analyses based on intramolecular interactions. A comprehensive study of a series of analogues described in this paper suggests that a unique sulfur-nitrogen nonbonding interaction in the N-pyrimidyl/pyridyl-2-thiazolamine moiety enable conformations that are essential for activity. Further, a SARs study around the N-pyrimidyl/pyridyl-2-thiazolamine core culminated in the discovery of compound 3g, which showed potent in vitro PAM activity for the M3 mAChR with excellent subtype selectivity. Compound 3g also showed a distinct pharmacological effect on isolated smooth muscle tissue from rat bladder and favorable pharmacokinetics profiles, suggesting its potential as a chemical tool for probing the M3 mAChR in further research.

Discovery of novel scaffolds for γ-secretase modulators without an arylimidazole moiety.

Gamma-secretase modulators (GSMs) selectively inhibit the production of amyloid-ß 42 (Aß42) and may therefore be useful in the management of Alzheimer's disease. Most heterocyclic GSMs that are not derived from nonsteroidal anti-inflammatory drugs contain an arylimidazole moiety that potentially inhibits cytochrome P450 (CYP) activity. Here, we discovered imidazopyridine derivatives that represent a new class of scaffold for GSMs, which do not have a strongly basic end group such as arylimidazole. High-throughput screening identified 2-methyl-8-[(2-methylbenzyl)oxy]-3-(pyridin-4-yl)imidazo[1,2-a]pyridine (3a), which inhibited the cellular production of Aß42 (IC50 = 7.1 µM) without changing total production of Aß. Structural optimization of this series of compounds identified 5-[8-(benzyloxy)-2-methylimidazo[1,2-a]pyridin-3-yl]-2-ethylisoindolin-1-one (3m) as a potent inhibitor of Aß42 (IC50 = 0.39 µM) but not CYP3A4. Further, 3m demonstrated a sustained pharmacokinetic profile in mice and sufficiently penetrated the brain.

Prediction of Ligand Binding Affinity to Target Proteins by Molecular Mechanics Theoretical Calculation.

Accurate estimation of ligand-receptor binding affinity is indispensable for computer-assisted drug discovery and structure-based drug design. Many computational scoring functions for estimating binding affinity have been proposed. Every scoring function reported so far, however, has strengths and weaknesses depending on the chemical properties of ligands and the feature of the binding site of the receptor. Hence, potential functions that can be used for many kinds of target proteins are required. In this work, we developed a software program based on Morse-type potential functions that enables evaluation of binding affinity and geometry optimization. Eight different kinds of proteins were used as test data, and ligand chemicals for which the binding pose to the protein and inhibitory constant are known were selected for evaluation. The calculated binding score and the experimentally measured inhibitory constant showed good compatibilities for six target proteins but poor correlation for one target. These compatibilities were compared with the results obtained by using two other software programs. The comparison suggested that the performance of the software developed in this work is good. Since the software can be handled in a computer facility with a many-core system, the software will be effective for search for an active compound from a chemical database and for assistance in chemical modification of the active compound in the pharmaceutical research field.

Structural insight of DNA topoisomerases I from camptothecin-producing plants revealed by molecular dynamics simulations.

DNA topoisomerase I (Top1) catalyzes changes in DNA topology by cleaving and rejoining one strand of the double stranded (ds)DNA. Eukaryotic Top1s are the cellular target of the plant-derived anticancer indole alkaloid camptothecin (CPT), which reversibly stabilizes the Top1-dsDNA complex. However, CPT-producing plants, including Camptotheca acuminata, Ophiorrhiza pumila and Ophiorrhiza liukiuensis, are highly resistant to CPT because they possess point-mutated Top1. Here, the adaptive convergent evolution is reported between CPT production ability and mutations in their Top1, as a universal resistance mechanism found in all tested CPT-producing plants. This includes Nothapodytes nimmoniana, one of the major sources of CPT. To obtain a structural insight of the resistance mechanism, molecular dynamics simulations of CPT- resistant and -sensitive plant Top1s complexed with dsDNA and topotecan (a CPT derivative) were performed, these being compared to that for the CPT-sensitive human Top1. As a result, two mutations, Val617Gly and Asp710Gly, were identified in O. pumila Top1 and C. acuminata Top1, respectively. The substitutions at these two positions, surprisingly, are the same as those found in a CPT derivative-resistant human colon adenocarcinoma cell line. The results also demonstrated an increased linker flexibility of the CPT-resistant Top1, providing an additional explanation for the resistance mechanism found in CPT-producing plants. These mutations could reflect the long evolutionary adaptation of CPT-producing plant Top1s to confer a higher degree of resistance.

Mechanism of drug resistance of hemagglutinin of influenza virus and potent scaffolds inhibiting its function.

Highly pathogenic influenza viruses have become a global threat to humans. It is important to select an effective therapeutic option suitable for the subtypes in an epidemic or pandemic. To increase the options, the development of novel antiviral agents acting on targets different from those of the currently approved drugs is required. In this study, we performed molecular dynamics simulations on a spike protein on the viral envelop, hemagglutinin, for the wild-type and three kinds of mutants using a model system consisting of a trimeric hemagglutinin complex, viral lipid membrane, solvation waters, and ions. A natural product, stachyflin, which shows a high level of antiviral activity specific to some subtypes of influenza viruses, was examined on binding to the wild-type hemagglutinin by docking simulation. The compound potency of stachyflin is, however, easily lost due to resistant mutations. From a comparison of simulation results between the wild-type and the resistant mutants, the reason for the drug resistance of hemagglutinin was clarified. Next, 8 compounds were selected from a chemical database by in silico screening, considering the findings from the simulations. Inhibitory activities to suppress the proliferation of influenza virus were measured by cell-based antiviral assays, and two chemical scaffolds were found to be potent for an inhibitor. More than 30 derivatives bearing either of these two chemical scaffolds were synthesized, and cell culture assays were carried out to evaluate the compound potency. Several derivatives displayed a high compound potency, and 50% effective concentrations of two synthesized compounds were below 1 µM.

Peptide HIV-1 integrase inhibitors from HIV-1 gene products.

Anti-HIV peptides with inhibitory activity against HIV-1 integrase (IN) have been found in overlapping peptide libraries derived from HIV-1 gene products. In a strand transfer assay using IN, inhibitory active peptides with certain sequential motifs related to Vpr- and Env-derived peptides were found. The addition of an octa-arginyl group to the inhibitory peptides caused a remarkable inhibition of the strand transfer and 3'-end-processing reactions catalyzed by IN and significant inhibition against HIV replication.

High-performance drug discovery: computational screening by combining docking and molecular dynamics simulations.

Virtual compound screening using molecular docking is widely used in the discovery of new lead compounds for drug design. However, this method is not completely reliable and therefore unsatisfactory. In this study, we used massive molecular dynamics simulations of protein-ligand conformations obtained by molecular docking in order to improve the enrichment performance of molecular docking. Our screening approach employed the molecular mechanics/Poisson-Boltzmann and surface area method to estimate the binding free energies. For the top-ranking 1,000 compounds obtained by docking to a target protein, approximately 6,000 molecular dynamics simulations were performed using multiple docking poses in about a week. As a result, the enrichment performance of the top 100 compounds by our approach was improved by 1.6-4.0 times that of the enrichment performance of molecular dockings. This result indicates that the application of molecular dynamics simulations to virtual screening for lead discovery is both effective and practical. However, further optimization of the computational protocols is required for screening various target proteins.

Derivatives of 5-nitro-furan-2-carboxylic acid carbamoylmethyl ester inhibit RNase H activity associated with HIV-1 reverse transcriptase.

The RNase H activity associated with human immunodeficiency virus type 1 (HIV-1) is an attractive target for an antiretroviral drug development. We screened 20000 small-molecular-weight compounds for RNase H inhibitors and identified a novel RNase H-inhibiting structure characterized by a 5-nitro-furan-2-carboxylic acid carbamoylmethyl ester (NACME) moiety. Two NACME derivatives, 5-nitro-furan-2-carboxylic acid adamantan-1-carbamoylmethyl ester (compound 1) and 5-nitro-furan-2-carboxylic acid [[4-(4-bromo-phenyl)-thiazol-2-yl]-(tetrahydro-furan-2-ylmethyl)-carbamoyl]-methyl ester (compound 2), effectively blocked HIV-1 and MLV RT-associated RNase H activities with IC(50)s of 3-30 microM but had little effect on bacterial RNase H activity in vitro. Additionally, 20-25 microM compound 2 effectively inhibited HIV-1 replication. An in silico docking simulation indicated that the conserved His539 residue, and two metal ions in the RNase H catalytic center are involved in RNase H inhibition by NACME derivatives. Taken together, these data suggest that NACME derivatives may be potent lead compounds for development of a novel class of antiretroviral drugs.

Ab initio protein structure prediction with force field parameters derived from water-phase quantum chemical calculation.

Molecular dynamics (MD) simulations are extensively used in the study of the structures and functions of proteins. Ab initio protein structure prediction is one of the most important subjects in computational biology, and many trials have been performed using MD simulation so far. Since the results of MD simulations largely depend on the force field, reliable force field parameters are indispensable for the success of MD simulation. In this work, we have modified atom charges in a standard force field on the basis of water-phase quantum chemical calculations. The modified force field turned out appropriate for ab initio protein structure prediction by the MD simulation with the generalized Born method. Detailed analysis was performed in terms of the conformational stability of amino acid residues, the stability of secondary structure of proteins, and the accuracy for prediction of protein tertiary structure, comparing the modified force field with a standard one. The energy balance between alpha-helix and beta-sheet structures was significantly improved by the modification of charge parameters.