Industry-scale application and evaluation of deep learning for drug target prediction.

Artificial intelligence (AI) is undergoing a revolution thanks to the breakthroughs of machine learning algorithms in computer vision, speech recognition, natural language processing and generative modelling. Recent works on publicly available pharmaceutical data showed that AI methods are highly promising for Drug Target prediction. However, the quality of public data might be different than that of industry data due to different labs reporting measurements, different measurement techniques, fewer samples and less diverse and specialized assays. As part of a European funded project (ExCAPE), that brought together expertise from pharmaceutical industry, machine learning, and high-performance computing, we investigated how well machine learning models obtained from public data can be transferred to internal pharmaceutical industry data. Our results show that machine learning models trained on public data can indeed maintain their predictive power to a large degree when applied to industry data. Moreover, we observed that deep learning derived machine learning models outperformed comparable models, which were trained by other machine learning algorithms, when applied to internal pharmaceutical company datasets. To our knowledge, this is the first large-scale study evaluating the potential of machine learning and especially deep learning directly at the level of industry-scale settings and moreover investigating the transferability of publicly learned target prediction models towards industrial bioactivity prediction pipelines.

Interaction Between the Rubidium Cation and [2.2.2]Paracyclophane: Experimental and Theoretical Study.

By means of electrospray ionization mass spectrometry (ESI-MS), it was evidenced experimentally that the rubidium cation (Rb+) reacts with the electroneutral [2.2.2]paracyclophane ligand (C24H24) to form the cationic complex [Rb(C24H24)]+. Moreover, applying quantum chemical calculations, the most probable conformation of the proven [Rb(C24H24)]+ complex was solved. In the complex [Rb(C24H24)]+ having a symmetry very close to C3, the "central" cation Rb+, fully located in the cavity of the parent [2.2.2]paracyclophane ligand, is coordinated to all three benzene rings of [2.2.2]paracyclophane via cation-π interaction. Finally, the binding energy, E(int), of the considered cation-π complex [Rb(C24H24)]+ was evaluated as -99.3 kJ/mol, confirming the formation of this fascinating complex species as well. This means that the [2.2.2]paracyclophane ligand can be considered as a receptor for the rubidium cation in the gas phase.

Affinity capillary electrophoresis and density functional theory study of noncovalent interactions of cyclic peptide [Gly6 ]-antamanide with small cations.

ACE and density functional theory were employed to study the noncovalent interactions of cyclic decapeptide glycine-6-antamanide ([Gly6 ]AA), synthetic derivative of native antamanide (AA) peptide from the deadly poisonous fungus Amanita phalloides, with small cations (Li+ , Rb+ , Cs+ , NH4+ , and Ca2+ ) in methanol. The strength of these interactions was quantified by the apparent stability constants of the appropriate complexes determined by ACE. The stability constants were calculated using the nonlinear regression analysis of the dependence of the effective electrophoretic mobility of [Gly6 ]AA on the concentration of the above ions in the BGE (methanolic solution of 20 mM chloroacetic acid, 10 mM Tris, pHMeOH 7.8, containing 0-70 mM concentrations of the above ions added in the form of chlorides). Prior to stability constant calculation, the effective mobilities measured at actual temperature inside the capillary and at variable ionic strength of the BGEs were corrected to the values corresponding to the reference temperature of 25°C and to the constant ionic strength of 10 mM. From the above ions, Rb+ and Cs+ cations interacted weakly with [Gly6 ]AA but no interactions of [Gly6 ]AA with univalent Li+ and NH4+ ions and divalent Ca2+ ion were observed. The apparent stability constants of [Gly6 ]AA-Rb+ and [Gly6 ]AA-Cs+ complexes were found to be equal to 13 ± 4 and 22 ± 3 L/mol, respectively. The structural characteristics of these complexes, such as position of the Rb+ and Cs+ ions in the cavity of the [Gly6 ]AA molecule and the interatomic distances within these complexes, were obtained by the density functional theory calculations.

Affinity capillary electrophoresis and quantum mechanical calculations applied to investigation of [Gly6 ]-antamanide binding with sodium and potassium ions.

ACE in a free solution and quantum mechanical density functional theory have been applied to the investigation of interactions of glycine-6-antamanide ([Gly6 ]AA), a synthetic derivative of cyclic decapeptide antamanide isolated from the highly poisonous mushroom Amanita phalloides, with sodium or potassium ions in methanol. First, from the dependence of effective electrophoretic mobility of [Gly6 ]AA on Na+ or K+ ions concentration in the BGE (methanolic solution of 20 mM chloroacetic acid, 10 mM Tris, pHMeOH 7.8, containing 0-50 mM NaCl or 0-40 mM KCl), the apparent binding (stability) constants of [Gly6 ]AA-Na+ and [Gly6 ]AA-K+ complexes were evaluated as 26 ± 1 and 14 ± 1 L/mol, respectively. The employed ACE method included correction of the effective mobilities measured at ambient temperature and at variable ionic strength of the BGEs to the mobilities related to the reference temperature 25°C and to the constant ionic strength 10 mM. Second, the interaction energies of the [Gly6 ]AA-Na+ and [Gly6 ]AA-K+ complexes (-466.3 and -345.2 kJ/mol, respectively) and the structural details of these complexes, such as position of the Na+ and K+ ions in the cavity of the [Gly6 ]AA molecule and the interatomic distances within these complexes, were determined by the density functional theory calculations.

Synthesis of inherently chiral calixarenes via direct mercuration of the partial cone conformation.

Direct mercuration of calix[4]arene immobilized in the partial cone conformation led to the meta-substituted isomer which was subsequently subjected to Pd-catalysed coupling (C-H activation) with the neighbouring aromatic subunit. Regioselective mercuration thus enabled access to a novel type of inherently chiral calixarenes with a highly distorted cavity potentially applicable to the design of new chiral receptors.

Regioselective halogenation of thiacalix[4]arenes in the cone and 1,3-alternate conformations.

Monohalogenation of thiacalix[4]arene in the cone conformation gave either the meta- or para-substituted isomers depending on the halogen and reaction conditions used. Surprisingly, the same reaction with the 1,3-alternate conformer led only to the meta isomer. This is the first example of such a conformation-dependent regioselectivity in calixarene/thiacalixarene chemistry. As the halogen-substituted calixarenes are useful synthetic intermediates, this provided the unique opportunity to functionalize the basic skeleton at two different positions.

Novel tacrine/acridine anticholinesterase inhibitors with piperazine and thiourea linkers.

A new series of substituted tacrine/acridine and tacrine/tacrine dimers with aliphatic or alkylene-thiourea linkers was synthesized and the potential of these compounds as novel human acetylcholinesterase (hAChE) and human butyrylcholinesterase (hBChE) inhibitors with nanomolar inhibition activity was evaluated. The most potent AChE inhibitor was found to be homodimeric tacrine derivative 14a, which demonstrated an IC50 value of 2 nM; this value indicates an activity rate which is 250-times higher than that of tacrine 1 and 7500-times higher than 7-MEOTA 15, the compounds which were used as standards in the study. IC50 values of derivatives 1, 9, 10, 14b and 15 were compared with the dissociation constants of the enzyme-inhibitor complex, Ki1, and the enzyme-substrate-inhibitor complex, Ki2, for. A dual binding site is presumed for the synthesized compounds which possess two tacrines or tacrine and acridine as terminal moieties show evidence of dual site binding. DFT calculations of theoretical desolvation free energies, ΔΔGtheor, and docking studies elucidate these suggestions in more detail.

Evidence for the cyclic CN2 carbene in solution.

Diazirinylidene (c-CN2) is formally the simplest of the N-heterocyclic carbenes. The intermediacy of this elusive species in the fragmentation of butyl 3-bromodiazirine-3-carboxylate (1a) with pent-4-en-1-ols and their sodium alkoxides in DMF is supported by the formation of 2-oxabicyclo[4.1.0]heptanes and dipentenoxymethanes. These products result from an intramolecular [2 + 1] cycloaddition and O-H insertion, respectively, of pentenoxymethylenes suggested to originate from the reaction of the electrophilic c-CN2 with an alkoxide ion. The reaction of 1a with primary or secondary amines in methanol affords the corresponding 3-bromodiazirine-3-carboxamides.

Dimercuration of calix[4]arenes: novel substitution pattern in calixarene chemistry.

A mercuration reaction of tetrapropoxycalix[4]arene immobilized in the cone conformation gave a mixture of two dimercurated products (meta,meta and meta,para) in approximately a 1:1 ratio. Both regioisomers represent inherently chiral compounds, which makes them very attractive for design of novel receptors. As demonstrated by Pd-catalyzed arylation, the different reactivity of HgCl functions in the meta,para-disubstituted isomer opens the door for regioselective introductions of two different functional groups to achieve a substitution pattern so far unknown in calixarene chemistry.

Unprecedented meta-substitution of calixarenes: direct way to inherently chiral derivatives.

Electrophilic aromatic substitution in the calix[n]arene series is a well-established procedure leading exclusively to para-substituted derivatives. An unprecedented regioselectivity of the mercuration reaction leading to the meta-substituted calix[4]arenes is described. These compounds represent a new type of substitution pattern in classical calixarene chemistry and open the door for the straightforward synthesis of inherently chiral receptors based on calixarenes.

S-alkylation of thiacalixarenes: how the regio- and stereoselectivities depend on the starting conformation.

S-alkylation of all four thiacalix[4]arene conformations was accomplished using alkyl triflates. The corresponding sulfonium salts are formed in a highly regio- and stereoselective manner depending on the conformation used. Interestingly, only mono- or disubstituted sulfonium salts can be prepared. Although many regio- and stereoisomers are theoretically possible, only one dialkylated cone and 1,2-alternate derivatives were formed, while only a single isomer of monoalkylated partial cone and 1,3-alternate were isolated. The combination of experimental results with the quantum-chemical approach using the B3LYP/6-311G(d,p) method resulted in the elucidation of the rules governing the regio- and stereochemical outcomes of the alkylation reactions. All S-alkylated compounds represent a novel type of substitution pattern in calixarene chemistry showing the wide-ranging possibility of thiacalixarene skeleton modifications.

Meta nitration of thiacalixarenes.

Nitration of thiacalix[4]arene, immobilized in the 1,3-alternate conformation, leads regioselectively to meta-substituted products. Depending on the reaction conditions, mono- and dinitro-derivatives can be isolated in acceptable yields. This unique substitution pattern is inaccessible in classical calixarene chemistry, and yields inherently chiral compounds, which makes thiacalixarenes very attractive as building blocks or molecular scaffolds.

Uncommon regioselectivity in thiacalix[4]arene formylation.

To reveal the alternative ways for upper-rim thiacalixarene derivatization, the formylation reactions (Gross and/or Duff conditions) of the corresponding tetrapropoxythiacalix[4]arene immobilized in the 1,3-alternate conformation were systematically studied. Surprisingly, albeit using an excess of the formylation agent, only two formyl groups were introduced exclusively into the meta positions of thiacalixarene skeleton. Unexpected regioselectivity of these reactions opens the door for a unique substitution pattern in thiacalixarene chemistry. The formation of meta-substituted aldehydes is another illustration showing remarkably different reactivity of the thiacalix[4]arene system compared with that of a classical calyx[4]arene analogue.

Spectroscopic, structural and theoretical studies of novel, potentially cytotoxic 4-acridonecarboxamide imines.

Ten novel, potentially intercalating 4-acridonecarboxamide azomethines and ketimines have been prepared by the condensation reaction of 9-oxo-9,10-dihydroacridine-4-carboxylic acid hydrazide with various aldehydes and ketones. The structures of the compounds were characterized spectroscopically by NMR ((1)H, (13)C, (15)N nuclei and 2D experiments), UV-vis, IR and fluorescence methods and by quantum chemical calculations using DFT at the B3LYP/6-311+G(d,p) level of theory and semiempirical ZINDO and AM1 methods. NMR chemical shift variations for C-4' were assessed due to changes in the polarizability of the imine C(4')=N(3') bond rather than direct inductive effects arising from the C-4' substituents. In concert with this was the reversed order observed for the N-3' chemical shifts with DFT-calculated atomic charges confirming the bond polarization. Both intra- and intermolecular hydrogen bonds between the acridone NH hydrogen and the amidic carbonyl oxygen were found to exist by FT-IR spectroscopy. Quantum chemical calculations were used to evaluate the configurational, tautomeric, conformational and hydrogen bonding states of the molecules as well as predict the NMR and IR data. The hypsochromic shifts observed in the UV-vis spectra upon changing from m-cresol to DMA, DMF or methanol were evaluated in terms of solvent polarity (giving rise to solvated excited state destabilization) and solvent aromaticity (giving rise to solvated excited state stabilization). The fluorescence of the compounds were modest, except for the 2,6-dichloro derivative, with respect to 9-isothiocyanatoacridine.

Interaction of two functional groups through the benzene ring: theory and experiment.

Energies of 132 benzene para bis-derivatives calculated within the framework of the density functional theory at the level B3LYP/6-311+G(d,p)//B3LYP/6-311+G(d,p) were used for correlations of two types. Correlation with the experimental enthalpies of formation clearly revealed that the published experimental data are generally not dependable and may be loaded with errors of more than 10 kJ mol(-1). On the other hand, the calculated relative energies are biased so that the interaction of the two substituents is systematically overestimated. This shortcoming was insignificant for our correlations of the second type, in which the interaction of substituents expressed in terms of isodesmic reactions was analyzed depending on the effects of inductive and resonance. The results depended strongly on the character of substituents. When one substituent is an electron donor and the other is an acceptor, the inductive-resonance model works and the classical resonance picture is adequate. With two acceptor substituents, this model is still acceptable with lower precision (as crossed conjugation), but with two donors it fails completely and may be acceptable only for a much restricted subclass of strong donors. Many correlations described in the literature must be viewed with great caution when they are based only on a relatively small number of data, in which substituents of different types are not represented in a comparable number.

Molecular structure of guanine-quartet supramolecular assemblies in a gel-state based on a DFT calculation of infrared and vibrational circular dichroism spectra.

The infrared (IR) and vibrational circular dichroism (VCD) spectra of guanosine-5'-hydrazide ( G-1), a powerful hydrogelator, have been measured and analyzed on the basis of ab initio modeling. B3LYP/6-31G** DFT calculations predict that G-1, forming a clear solution in deuterated DMSO, is present in monomeric form in this solvent, whereas strong gelation in a phosphate buffer is due to the formation of a guanine-quartet structure, ( G-1)4, in which the four G-1 are linked by hydrogen-bonded guanine moieties and stabilized by an alkali metal cation. The B3LYP/6-31G** IR and VCD spectra of the nearly planar G-quartet, whose structure is slightly distorted from the C4h symmetry, in which the G-bases interact via four Hoogsteen-type hydrogen bonds and a sodium cation is positioned in the middle of the G-quartet, are in very good agreement with the experimental spectra, indicating that this structure is the predominant structure in the gel state. The geometric parameters are discussed. This study is the first to use IR and VCD spectroscopies coupled with DFT calculations to elucidate the structure of a supramolecular species in a gel state and shows the VCD spectroscopy as a powerful method for investigating the structure of complex supramolecular self-assemblies where the use of other structural methods is limited.

Regioselectivity and tautomerism of novel five-membered ring nitrogen heterocycles formed via cyclocondensation of acylthiosemicarbazides.

A series of 1-acyl-4-phenyl/(acridin-9-yl)thiosemicarbazides 3, including four new compounds, were prepared in order to study substituent effects on cyclization reactions with oxalyl chloride (producing imidazolidine-4,5-diones 4), dimethyl acetylenedicarboxylate (to give thiazolidin-4-ones 7 and 8) and autocondensation under alkaline conditions (to yield 1,2,4-triazoles 9). A positional isomer, 10 of compound 3f was also prepared. Altogether, twenty new compounds characterized and identified by IR, UV,1H, 13C and 2D NMR and quantum chemical calculations are described. The tautomerism of the products and regioselectivity of the reactions were evaluated. Compounds 3f-h,3h.2HCl, 7b,d and 10 were screened for cytotoxic activity against the L1210 leukemia cell line and all compounds, except for 3f, exhibited promising inhibitions of cell growth.

Steric effects of polar substituents evaluated in terms of energy by means of isodesmic reactions.

Steric effects of various polar and some charged groups were estimated on sterically crowded cyclopropane cis-1,2-bis derivatives 2 or 3, in which the variable substituent is in the proximity of a t-butyl group or of a methyl group. The steric energy was evaluated with reference to the pertinent mono derivatives, that is as reaction energy of an isodesmic reaction, in which the crowded compound is formally synthesized from simple derivatives. Energies were calculated within the framework of the density functional theory at level B3LYP/6-311+G(d,p)//B3LYP/6-311+G(d,p) for 11 dipolar and 5 charged substituents. Interaction of charged substituents is not only steric (destabilizing) but also inductive (stabilizing). The steric effects evaluated in this way differ distinctly from the standard steric constants derived purely from the van der Waals radii of the substituents.

Formation and temperature stability of G-quadruplex structures studied by electronic and vibrational circular dichroism spectroscopy combined with ab initio calculations.

Variations in the structure of d(GGGA)(5) oligonucleotide in the presence of Li(+), Na(+), and K(+) ions and its temperature stability were studied using electronic and vibrational circular dichroism, IR absorption, and ab initio calculations with the Becke 3-Lee-Yang-Parr functional at the 6-31G** level. The samples were characterized by nondenaturing gel electrophoresis. Oligonucleotide d(GGGA)(5) in the presence of Li(+) forms a nonplanar single tetramer, with angles of 102 degrees and 171 degrees between neighboring guanine bases. This tetramer changes its geometry at temperatures >50 degrees C, but does not form a quadruplex structure. In the presence of Na(+), the d(GGGA)(5) structure was optimized to almost planar tetramers with an angle of 177 degrees between neighboring guanines. The spectral results suggest that it stacks into a quadruplex helical structure. This quadruplex structure decayed to a single tetramer at temperatures >60 degrees C. The Hartree-Fock energies imply that d(GGGA)(5) prefers to form complexes with Na(+) rather than Li(+). The d(GGGA)(5) structure in the presence of monovalent ions is stabilized against thermal denaturation in the order Li(+) < Na(+) < K(+).

Steric effects of the alkyl groups: evaluation on isolated molecules by means of isodesmic reactions.

Several possible scales of steric effects of the alkyl groups were suggested on the basis of isodesmic model reactions, in which a sterically crowded compound is formally synthesized from simpler derivatives. The reaction energies were calculated within the framework of the density functional theory at the level B3LYP/6-311+G(d.p)//B3LYP/6-311+G(d.p) for 6 model systems and 7 various alkyl groups. The most important systems were cis-1,2-dialkylcyclopropanes 1 synthesized from two mono derivatives and sterically crowded derivatives of bicyclo[2.2.2]octane 2 with C(3) symmetry. The scales of steric effects evaluated from the two models were rather different: the first scale depended in effect only on the C atoms in the alpha and beta positions and the effects were almost equal for all primary alkyls. The second scale depended also on the gamma position and the effect of the CH(2)-t-Bu group was much greater than that of the ethyl group. Any relationship between various systems was found rarely, only in the case of very similar reaction series; even in such cases the relationship was sometimes linear, sometimes distinctly curvilinear. It is concluded that any universal scale of steric effects is in principle not possible since these effects depend specifically on the surroundings of the substituent in a particular reaction. Nevertheless, there is a similarity between various scales; a bulky group appears as bulky in any scale. Therefore, very rough correlations of steric effects are possible.