Improving (Q)SAR predictions by examining bias in the selection of compounds for experimental testing.

Existing data on structures and biological activities are limited and distributed unevenly across distinct molecular targets and chemical compounds. The question arises if these data represent an unbiased sample of the general population of chemical-biological interactions. To answer this question, we analyzed ChEMBL data for 87,583 molecules tested against 919 protein targets using supervised and unsupervised approaches. Hierarchical clustering of the Murcko frameworks generated using Chemistry Development Toolkit showed that the available data form a big diffuse cloud without apparent structure. In contrast hereto, PASS-based classifiers allowed prediction whether the compound had been tested against the particular molecular target, despite whether it was active or not. Thus, one may conclude that the selection of chemical compounds for testing against specific targets is biased, probably due to the influence of prior knowledge. We assessed the possibility to improve (Q)SAR predictions using this fact: PASS prediction of the interaction with the particular target for compounds predicted as tested against the target has significantly higher accuracy than for those predicted as untested (average ROC AUC are about 0.87 and 0.75, respectively). Thus, considering the existing bias in the data of the training set may increase the performance of virtual screening.

[Discovering new antiretroviral compounds in "Big Data" chemical space of the SAVI library]. / Poisk novykh antiretrovirusnykh soedinenii v khimicheskom prostranstve "bol'shikh dannykh" biblioteki SAVI.

Despite significant advances in the application of highly active antiretroviral therapy, the development of new drugs for the treatment of HIV infection remains an important task because the existing drugs do not provide a complete cure, cause serious side effects and lead to the emergence of resistance. In 2015, a consortium of American and European scientists and specialists launched a project to create the SAVI (Synthetically Accessible Virtual Inventory) library. Its 2016 version of over 283 million structures of new easily synthesizable organic molecules, each annotated with a proposed synthetic route, were generated in silico for the purpose of searching for safer and more potent pharmacological substances. We have developed an algorithm for comparing large chemical databases (DB) based on the representation of structural formulas in SMILES codes, and evaluated the possibility of detecting new antiretroviral compounds in the SAVI database. After analyzing the intersection of SAVI with 97 million structures of the PubChem database, we found that only a small part of the SAVI (~0.015%) is represented in PubChem, which indicates a significant novelty of this virtual library. However, among those structures, 632 compounds tested for anti-HIV activity were detected, 41 of which had the desired activity. Thus, our studies for the first time demonstrated that SAVI is a promising source for the search for new anti-HIV compounds.

Computer-aided design and discovery of protein-protein interaction inhibitors as agents for anti-HIV therapy.

Protein-protein interactions (PPI) are involved in most of the essential processes that occur in organisms. In recent years, PPI have become the object of increasing attention in drug discovery, particularly for anti-HIV drugs. Although the use of combinations of existing drugs, termed highly active antiretroviral therapy (HAART), has revolutionized the treatment of HIV/AIDS, problems with these agents, such as the rapid emergence of drug-resistant HIV-1 mutants and serious adverse effects, have highlighted the need for further discovery of new drugs and new targets. Numerous investigations have shown that PPI play a key role in the virus's life cycle and that blocking or modulating them has a significant therapeutic potential. Here we summarize the recent progress in computer-aided design of PPI inhibitors, mainly focusing on the selection of the drug targets (HIV enzymes and virus entry machinery) and the utilization of peptides and small molecules to prevent a variety of protein-protein interactions (viral-viral or viral-host) that play a vital role in the progression of HIV infection.

Internet resources integrating many small-molecule databases.

()New data, tools and services recently made available on the web server (http://cactus.nci.nih.gov) of the Computer-Aided Drug Design (CADD) Group, NCI, NIH, developed in the context of chemoinformatics and drug development work, are presented. These tools are designed for searching for structures in very large databases of small molecules. One of them is a web service-the Chemical Structure Look-up Service (CSLS)-for very rapid structure look-up in an aggregated collection of more than 80 databases comprising more than 27 million unique structures at the time of this writing. CSLS contains pointers to the entries in toxicology-related databases, catalogues of commercially available samples, drugs, assay results data sets, and databases in several other categories. CSLS allows the user to find out very rapidly in which one(s) of all these databases a given structure occurs independent of the representation of the input structure, by making use of InChIs as well as new CACTVS hashcode-based identifiers. These latter, calculable, identifiers are designed to take into account tautomerism, different resonance structures drawn for charged species, and presence of additional fragments. They make possible fine-tunable yet rapid compound identification and database overlap analyses in very large compound collections.

Conformationally constrained analogues of diacylglycerol. 18. The incorporation of a hydroxamate moiety into diacylglycerol-lactones reduces lipophilicity and helps discriminate between sn-1 and sn-2 binding modes to protein kinase C (PK-C). Implications for isozyme specificity.

An approach to reduce the log P in a series of diacylglycerol (DAG)-lactones known for their high binding affinity for protein kinase C (PK-C) is presented. Branched alkyl groups with reduced lipophilicity were selected and combined with the replacement of the ester or lactone oxygens by NH or NOH groups. Compound 6a with an isosteric N-hydroxyl amide arm represents the most potent and least lipophilic DAG analogue known to date.

Inhibition of (cytosine C5)-methyltransferase by oligonucleotides containing flexible (cyclopentane) and conformationally constrained (bicyclo[3.1.0]hexane) abasic sites.

Pseudorotationally locked sugar analogues based on bicyclo[3.1.0]-hexane templates were placed in DNA duplexes as abasic target sites in the M. HhaI recognition sequence. The binding affinity of the enzyme increases when the abasic site is constrained to the South conformation and decreases when it is constrained to the North conformation. A structural understanding of these differences is provided.

Comparison of the NCI open database with seven large chemical structural databases.

Eight large chemical databases have been analyzed and compared to each other. Central to this comparison is the open National Cancer Institute (NCI) database, consisting of approximately 250 000 structures. The other databases analyzed are the Available Chemicals Directory ("ACD," from MDL, release 1.99, 3D-version); the ChemACX ("ACX," from CamSoft, Version 4.5); the Maybridge Catalog and the Asinex database (both as distributed by CamSoft as part of ChemInfo 4.5); the Sigma-Aldrich Catalog (CD-ROM, 1999 Version); the World Drug Index ("WDI," Derwent, version 1999.03); and the organic part of the Cambridge Crystallographic Database ("CSD," from Cambridge Crystallographic Data Center, 1999 Version 5.18). The database properties analyzed are internal duplication rates; compounds unique to each database; cumulative occurrence of compounds in an increasing number of databases; overlap of identical compounds between two databases; similarity overlap; diversity; and others. The crystallographic database CSD and the WDI show somewhat less overlap with the other databases than those with each other. In particular the collections of commercial compounds and compilations of vendor catalogs have a substantial degree of overlap among each other. Still, no database is completely a subset of any other, and each appears to have its own niche and thus "raison d'être". The NCI database has by far the highest number of compounds that are unique to it. Approximately 200 000 of the NCI structures were not found in any of the other analyzed databases.

An optimized protein kinase C activating diacylglycerol combining high binding affinity (Ki) with reduced lipophilicity (log P).

A small, focused combinatorial library encompassing all possible permutations of acyl branched alkyl chains-small and large, saturated and unsaturated-was generated from the active diacylglycerol enantiomer (S-DAG) to help identify the analogue with the highest binding affinity (lowest Ki) for protein kinase C (PK-C) combined with the minimum lipophilicity (log P). The selected ligand (3B) activated PK-C more effectively than sn-1,2-dioctanoylglycerol (diC8) despite being 1.4 log units more hydrophilic. Compound 3B indeed represents the most potent, hydrophilic DAG ligand to date. With the help of a green fluorescent protein (GFP)-tagged PK-Calpha, 3B was able to translocate the full length protein to the membrane with an optimal dose of 100 microM in CHO-K1 cells, while diC8 failed to achieve translocation even at doses 3-fold higher. Molecular modeling of 3B into an empty C1b domain of PK-Cdelta clearly showed the existence of a preferred binding orientation. In addition, molecular dynamic simulations suggest that binding discrimination could result from a favorable van der Waals (VDW) interaction between the large, branched sn-1 acyl group of 3B and the aromatic rings of Trp252 (PK-Cdelta) or Tyr252 (PK-Calpha). The DAG analogue of 3B in which the acyl groups are reversed (2C) showed a decrease in binding affinity reflecting the capacity of PK-C to effectively discriminate between alternative orientations of the acyl chains.

Identification of HIV-1 integrase inhibitors via three-dimensional database searching using ASV and HIV-1 integrases as targets.

Integration of viral DNA into the host cell genome is a critical step in the life cycle of HIV. This essential reaction is catalyzed by integrase (IN) through two steps, 3'-processing and DNA strand transfer. Integrase is an attractive target for drug design because there is no known cellular analogue and integration is essential for successful replication of HIV. A computational three-dimensional (3-D) database search was used to identify novel HIV-1 integrase inhibitors. Starting from the previously identified Y3 (4-acetylamino-5-hydroxynaphthalene-2,7-disulfonic acid) binding site on the avian sarcoma virus integrase (ASV IN), a preliminary search of all compounds in the nonproprietary, open part of the National Cancer Institute 3-D database yielded a collection of 3100 compounds. A more rigorous scoring method was used to rescreen the 3100 compounds against both ASV IN and HIV-1 IN. Twenty-two of those compounds were selected for inhibition assays against HIV-1 IN. Thirteen of the 22 showed inhibitory activity against HIV-1 IN at concentrations less than 200 microM and three of them showed antiviral activities in HIV-1 infected CEM cells with effective concentrations (EC50) ranging from 0.8 to 200 microM. Analysis of the computer-generated binding modes of the active compounds to HIV-1 IN showed that simultaneous interaction with the Y3 site and the catalytic site is possible. In addition, interactions between the active compounds and the flexible loop involved in the binding of DNA by IN are indicated to occur. The structural details and the unique binding motif between the HIV-1 IN and its inhibitors identified in the present work may contribute to the future development of IN inhibitors.

Interactions of conformationally biased north and south 2'-fluoro-2', 3'-dideoxynucleoside 5'-triphosphates with the active site of HIV-1 reverse transcriptase.

Molecular dynamics simulations of a ternary complex of HIV-1 reverse transcriptase (RT), double-stranded DNA, and bound dideoxynucleoside-5'-triphosphate (RT-DNA-ddNTP), utilizing the ddNTPs ddATP, betaFddATP, and alphaFddATP, explain the experimentally observed order of potency of these 5'-triphosphates as inhibitors of RT: ddATP > betaFddATP > alphaFddATP. On the basis of RT's known preference to bind the incoming dNTP (or ddNTP) with a north conformation at the polymerase site, alphaFddATP, which in solution prefers almost exclusively a north conformation, was predicted to be the most potent inhibitor. However, Tyr115, which appears to function as a steric gate to preclude the binding of ribonucleoside 5'-triphosphates, prevents the effective binding of alphaFddATP in its preferred north conformation. The south-biased betaFddATP, while able to bind to RT without hindrance by Tyr115, has to pay a high energy penalty to be flipped to the active north conformation at the polymerase site. Finally, the more flexible and less conformationally biased ddATP is able to switch to a north conformation at the RT site with a smaller energy penalty than betaFddATP. These results highlight the opposite conformational preferences of HIV-1 RT for alphaFddATP and betaFddATP and help establish conformational guidelines for optimal binding at the polymerase site of this enzyme.

Adenosine deaminase prefers a distinct sugar ring conformation for binding and catalysis: kinetic and structural studies.

Several recent X-ray crystal structures of adenosine deaminase (ADA) in complex with various adenosine surrogates have illustrated the preferred mode of substrate binding for this enzyme. To define more specific structural details of substrate preferences for binding and catalysis, we have studied the ADA binding efficiencies and deamination kinetics of several synthetic adenosine analogues in which the furanosyl ring is biased toward a particular conformation. NMR solution studies and pseudorotational analyses were used to ascertain the preferred furanose ring puckers (P, nu(MAX)) and rotamer distributions (chi and gamma) of the nucleoside analogues. It was shown that derivatives which are biased toward a "Northern" (3'-endo, N) sugar ring pucker were deaminated up to 65-fold faster and bound more tightly to the enzyme than those that preferred a "Southern" (2'-endo, S) conformation. This behavior, however, could be modulated by other structural factors. Similarly, purine riboside inhibitors of ADA that prefer the N hemisphere were more potent inhibitors than S analogues. These binding propensities were corroborated by detailed molecular modeling studies. Docking of both N- and S-type analogues into the ADA crystal structure coordinates showed that N-type substrates formed a stable complex with ADA, whereas for S-type substrates, it was necessary for the sugar pucker to adjust to a 3'-endo (N-type) conformation to remain in the ADA substrate binding site. These data outline the intricate structural details for optimum binding in the catalytic cleft of ADA.

Conformationally restricted nucleosides. The reaction of adenosine deaminase with substrates built on a bicyclo[3.1.0]hexane template.

Adenosine deaminase (ADA) can discriminate between two distinct (North and South), conformationally rigid substrate conformers. (N)-methanocarba-2'dA (4) is deaminated 100 times faster than the antipodal (S)-methanocarba-2'dA (5), whereas a non-rigid analogue, aristeromycin (6), is deaminated at an intermediate rate. These results are in agreement with crystallographic data from ADA-ribonucleoside complexes showing the furanose ring of the bound purine in a C3'-endo (North) conformation. The data presented here suggests that 4 and 5 are useful probes to ascertain conformational preferences by purine metabolizing enzymes.

Mutation of a conserved serine residue in a quinolone-resistant type II topoisomerase alters the enzyme-DNA and drug interactions.

A Ser740 --> Trp mutation in yeast topoisomerase II (top2) and of the equivalent Ser83 in gyrase results in resistance to quinolones and confers hypersensitivity to etoposide (VP-16). We characterized the cleavage complexes induced by the top2(S740W) in the human c-myc gene. In addition to resistance to the fluoroquinolone CP-115,953, top2(S740W) induced novel DNA cleavage sites in the presence of VP-16, azatoxin, amsacrine, and mitoxantrone. Analysis of the VP-16 sites indicated that the changes in the cleavage pattern were reflected by alterations in base preference. C at position -2 and G at position +6 were observed for the top2(S740W) in addition to the previously reported C-1 and G+5 for the wild-type top2. The VP-16-induced top2(S740W) cleavage complexes were also more stable. The most stable sites had strong preference for C-1, whereas the most reversible sites showed no base preference at positions -1 or -2. Different patterns of DNA cleavage were also observed in the absence of drug and in the presence of calcium. These results indicate that the Ser740 --> Trp mutation alters the DNA recognition of top2, enhances its DNA binding, and markedly affects its interactions with inhibitors. Thus, residue 740 of top2 appears critical for both DNA and drug interactions.

Computational chemistry on commodity-type computers.

A number of inexpensive computers were benchmarked with the ab initio program Gaussian 94, using both small standard test jobs and larger density functional (DFT) calculations. Several varieties of Pentium (x86) and Alpha CPU based systems were tested. Most of them were running under the open source code operating system Linux. They were compared with several workstations and supercomputers. The most powerful of today's commodity-type processors surpassed current supercomputers in speed. The choice of compilers and compilation options was often found to have a larger influence on job CPU times than details of the hardware. Especially on the x86 type machines, the jobs always ran faster the less memory (RAM) they were given. The fastest machine on a per-CPU basis was an Alpha/Linux system. For the DFT calculation, it was close to twice as fast as a Cray J90 supercomputer.

Pharmacophores in drug design and discovery.

Compounds containing a specific pharmacophore--the minimum structural features necessary for enzyme binding--can be retrieved from a database such as the National Cancer Institute repository by means of three-dimensional (3D) searching, which allows the retrieval of all compounds possessing a specified set of atoms with a given 3D geometry. The ways in which pharmacophores can be found and characterized and the details of the 3D searching methods are described. Elaboration of compounds found in such searches and their subsequent development as lead drugs is also discussed.

HIV-1 integrase pharmacophore: discovery of inhibitors through three-dimensional database searching.

Starting from a known inhibitor of human immunodeficiency virus type 1 (HIV-1) integrase (IN); caffeic acid phenethyl ester (CAPE), a putative three-point pharmacophore for binding of inhibitors to IN was derived. This pharmacophore was used to search the National Cancer Institute three-dimensional (3D) structural database. Out of the open, nonproprietary part of this database, comprising approximately 200000 compounds, 267 structures were found to match the pharmacophore in at least one conformation, and 60 of those were tested in an in vitro assay against HIV-1 IN. Out of these, 19 were found to inhibit both the 3'-processing and strand transfer of IN at micromolar concentrations. In order to test the validity of this pharmacophore, a small 3D database of 152 published IN inhibitors was built. A search in this database yielded a statistically significant correlation of the presence of this pharmacophore and the potency of the compounds. An automated pharmacophore identification procedure performed on this set of compounds provided additional support for the importance of this pharmacophore for binding of inhibitors to IN and hinted at a possible second pharmacophore. The role of aromatic moieties in the binding of ligands to HIV-1 IN through interactions with divalent metal cations, which are known to be necessary for activity of the enzyme, was explored in ab initio calculations.

Discovery of HIV-1 integrase inhibitors by pharmacophore searching.

Based upon a class of known HIV-1 integrase inhibitors, several pharmacophore models were proposed from molecular modeling studies and validated using a 3D database of 152, compounds for which integrase assay data are known. Using the most probable pharmacophore model as the query, the NCI 3D database of 206,876 compounds was searched, and 340 compounds that contain the pharmacophore query were identified. Twenty-nine of these compounds were selected and tested in the HIV-1 integrase assay. This led to the discovery of 10 novel, structurally diverse HIV-1 integrase inhibitors, four of which have an IC50 value less than 30 microM and are promising lead compounds for further HIV-1 integrase inhibitor development.

Depsides and depsidones as inhibitors of HIV-1 integrase: discovery of novel inhibitors through 3D database searching.

Seventeen lichen acids comprising despides, depsidones, and their synthetic derivatives have been examined for their inhibitory activity against HIV-1 integrase, and two pharmacophores associated with inhibition of this enzyme have been identified. A search of the NCI 3D database of approximately 200,000 structures yielded some 800 compounds which contain one or the other pharmacophore. Forty-two of these compounds were assayed for HIV-1 integrase inhibition, and of these, 27 had inhibitory IC50 values of less than 100 microM; 15 were below 50 microM. Several of these compounds were also examined for their activity against HIV-2 integrase and mammalian topoisomerase I.

Molecular modeling in the discovery of drug leads.

The National Cancer Institute of the U.S.A. maintains a repository of about 500,000 chemicals which it has tested at some time for anticancer activity. As new chemotherapeutic targets present themselves, methods have been developed by which this large database can be re-examined without resorting to expensive high-volume biological screening. Electronic screening, the method described in this paper, begins with the identification of a target enzyme. The pharmacophore used by the enzyme in binding to substrates is identified, and then all compounds in the database that contain the pharmacophore are found. These compounds are then further filtered, for example, by physical properties such as solubility, and the relatively small number of compounds that survive are submitted for biological testing. This use of a primary electronic screen in the search for ligands of protein kinase C is described. The screen is very fast, and the method is quite generally applicable to different enzymes.

Discovery of novel, non-peptide HIV-1 protease inhibitors by pharmacophore searching.

Fifteen novel non-peptide HIV-1 protease inhibitors were identified by flexible 3D database pharmacophore searching of the NCI DIS 3D database. The pharmacophore query used in the search was derived directly from the X-ray determined structures of protease/inhibitor complexes. These 15 inhibitors, belonging to nine different chemical classes, are promising leads for further development. The two best inhibitors found, NSC 32180, a "dimer" of 4-hydroxycoumarin, and NSC 117027, a "tetramer" of 2-hydroxy quinone, had ID50 values of 0.32 and 0.75 microM for HIV-1 protease inhibition, respectively, and two other inhibitors had ID50 values close to 1 microM. Among the potent inhibitors, NSC 158393 not only demonstrated activity against HIV-1 protease (ID50 1.7 microM) but also exhibited promising antiviral activity in HIV-1-infected CEM-SS cells (EC50 = 11.5 microM). Validation of the pharmacophore used in the search was accomplished by conformational analysis. The binding modes of the most potent inhibitor found in our studies, NSC 32180, were predicted employing docking and molecular dynamics techniques.