Finding Needles in a Haystack: Determining Key Molecular Descriptors Associated with the Blood-brain Barrier Entry of Chemical Compounds Using Machine Learning.

In this paper we used two sets of calculated molecular descriptors to predict blood-brain barrier (BBB) entry of a collection of 415 chemicals. The set of 579 descriptors were calculated by Schrodinger and TopoCluj software. Polly and Triplet software were used to calculate the second set of 198 descriptors. Following this, modelling and a two-deep, repeated external validation method was used for QSAR formulation. Results show that both sets of descriptors individually and their combination give models of reasonable prediction accuracy. We also uncover the effectiveness of a variable selection approach, by showing that for one of our descriptor sets, the top 5 % predictors in terms of random forest variable importance are able to provide a better performing model than the model with all predictors. The top influential descriptors indicate important aspects of molecular structural features that govern BBB entry of chemicals.

Cube-Related Corner Coalesced Nets.

Finite or periodic structures containing the cube motif can be synthesized and transformed into a variety of structures both at the theoretical and real, experimental level. The rhombellation topo-geometric operation may be used to transform the cube-shape into larger units and then build light (spongy) structures with larger voids. Hyper-clusters are polyhedral structures which nodes are polyhedral structures (the same or different ones). The paper presents some hypothetical spongy structures related to the cubic primitive pcu-net, with defects induced by cutting-off some atoms and/or bonds so that only corners are shared between two cubes. A diamondoid hyper-structure containing cube-coalesced corners was proposed for an eventual synthesis. The discussed structures are described in topological terms, particularly by sequential vertex connectivity and ring environment.

Binding Site and Potency Prediction of Teixobactin and other Lipid II Ligands by Statistical Base Scoring of Conformational Space Maps.

BACKGROUND: Lipid II, a peptidoglycan, is a precursor in bacterial cell synthesis. It has both hydrophilic and lipophilic properties. The molecule translocates a bacterial membrane to deliver and incorporate "building blocks" from disaccharide-pentapeptide into the peptidoglican wall. Lipid II is a valid antibiotic target. A receptor binding pocket may be occupied by a ligand in various plausible conformations, among which only few ones are energetically related to a biological activity in the physiological efficiency domain. This paper reports the mapping of the conformational space of Lipid II in its interaction with Teixobactin and other Lipid II ligands. METHODS: In order to study computationally the complex between Lipid II and ligands, a docking study was first carried on. Docking site was retrieved form literature. After docking, 5 ligand conformations and further 5 complexes (denoted 00 to 04) for each molecule were taken into account. For each structure, conformational studies were performed. Statistical analysis, conformational analysis and molecular dynamics based clustering were used to predict the potency of these compounds. A score for potency prediction was developed. RESULTS: Appling lipid II classification according to Lipid II conformational energy, a conformation of Teixobactin proved to be energetically favorable, followed by Oritravicin, Dalbavycin, Telvanicin, Teicoplamin and Vancomycin, respectively. Scoring of molecules according to cluster band and PCA produced the same result. Molecules classified according to standard deviations showed Dalbavycin as the most favorable conformation, followed by Teicoplamin, Telvanicin, Teixobactin, Oritravicin and Vancomycin, respectively. Total score showing best energetic efficiency of complex formation shows Teixobactin to have the best conformation (a score of 15 points) followed by Dalbavycin (14 points), Oritravicin (12v points), Telvanicin (10 points), Teicoplamin (9 points), Vancomycin (3 points). CONCLUSION: Statistical analysis of conformations can be used to predict the efficiency of ligand - target interaction and consecutively to find insight regarding ligand potency and postulate about favorable conformation of ligand and binding site. In this study it was shown that Teixobactin is more efficient in binding with Lipid II compared to Vancomycin, results confirmed by experimental data reported in literature.

Ligand Shaping in Induced Fit Docking of MraY Inhibitors. Polynomial Discriminant and Laplacian Operator as Biological Activity Descriptors.

Docking-i.e., interaction of a small molecule (ligand) with a proteic structure (receptor)-represents the ground of drug action mechanism of the vast majority of bioactive chemicals. Ligand and receptor accommodate their geometry and energy, within this interaction, in the benefit of receptor-ligand complex. In an induced fit docking, the structure of ligand is most susceptible to changes in topology and energy, comparative to the receptor. These changes can be described by manifold hypersurfaces, in terms of polynomial discriminant and Laplacian operator. Such topological surfaces were represented for each MraY (phospho-MurNAc-pentapeptide translocase) inhibitor, studied before and after docking with MraY. Binding affinities of all ligands were calculated by this procedure. For each ligand, Laplacian and polynomial discriminant were correlated with the ligand minimum inhibitory concentration (MIC) retrieved from literature. It was observed that MIC is correlated with Laplacian and polynomial discriminant.

Spongy Nanostructures.

Spongy structures are hollow-containing materials, encountered in natural or synthesized zeolites, spongy carbon, etc. The design and topological study of some hypothetical spongy nanostructures is presented in terms of map operations and genus calculation of their associated graphs, respectively. Among the discussed structures one remarks some novel spongy polyhedra that can evolve with 1-periodicity or radially, to provide multi-shell cages. Filling the space inside such complex nanostructures can be achieved by small fullerenes that self-arrange in aggregates with a well-defined geometry, of which energy trends to a minimal value. The way of space filling varies function of the dimension and shape of composing small fullerenes. An attempt of building and stability evaluating of several fullerene aggregates was made. The calculations were made at the HF, DFT and DFTB level of theory. The design of nanostructures was performed by our original software packages CVNET and Nano Studio.

Understanding the Action of Indolizines as Biologically Active Moieties: A Molecular Dynamics Study.

BACKGROUND: Indolizines represent a class of heteroaromatic compounds, of pharmacological importance, containing two condensed 5- and 6-memebered rings bridged by a nitrogen atom. Despite indolizine is an important medicinal moiety, a detailed view on the mechanism of action of biologically active indolizines is unavailable. OBJECTIVE: The study of ligand-enzyme affinity is of high interest; description of characteristics (energetic and geometric ones) of ligand binding to the active sites of an enzyme could be useful in understanding the action mechanism of a given ligand on the concerned enzyme. METHOD: After conducting a QSAR study, to predict IC50 (on 15-LO protein from soybeans) of indolizine derivatives and a docking study of indolizines on Beta lactamase and Nicotinamide phosphoribosyltransferase proteins [1], a molecular dynamics analysis was performed on one of the indolizine derivatives, complexed to the above proteins. RESULTS: The performed molecular dynamics study led to the identification of interactions responsible for the stabilization of complexes of the chosen ligand (i.e., indolizine derivative) with the considered enzymes and the specificity of the ligand interaction as well. The structural data and enthalpy values clearly indicate the differences in the behavior of ligand at the active sites of the three investigated enzymes. Among the studied proteins, the hydrophobicity of the active site of Nicotinamide phosphoribosyltransferase seems to be the main factor in promoting the interaction enzyme-ligand, much more manifested in this case, in comparison to the other two proteins Beta lactamase and Nicotinamide phosphoribosyltransferase. CONCLUSION: The present paper discusses a possible mechanism of interaction of an indolizine derivative with three enzyme proteins, providing information for future work in this topic.

Molecular Dynamic Studies of the Complex Polyethylenimine and Glucose Oxidase.

Glucose oxidase (GOx) is an enzyme produced by Aspergillus, Penicillium and other fungi species. It catalyzes the oxidation of ß-d-glucose (by the molecular oxygen or other molecules, like quinones, in a higher oxidation state) to form d-glucono-1,5-lactone, which hydrolyses spontaneously to produce gluconic acid. A coproduct of this enzymatic reaction is hydrogen peroxide (H2O2). GOx has found several commercial applications in chemical and pharmaceutical industries including novel biosensors that use the immobilized enzyme on different nanomaterials and/or polymers such as polyethylenimine (PEI). The problem of GOx immobilization on PEI is retaining the enzyme native activity despite its immobilization onto the polymer surface. Therefore, the molecular dynamic (MD) study of the PEI ligand (C14N8_07_B22) and the GOx enzyme (3QVR) was performed to examine the final complex PEI-GOx stabilization and the affinity of the PEI ligand to the docking sites of the GOx enzyme. The docking procedure showed two places/regions of major interaction of the protein with the polymer PEI: (LIG1) of -5.8 kcal/mol and (LIG2) of -4.5 kcal/mol located inside the enzyme and on its surface, respectively. The values of enthalpy for the PEI-enzyme complex, located inside of the protein (LIG1) and on its surface (LIG2) were computed. Docking also discovered domains of the GOx protein that exhibit no interactions with the ligand or have even repulsive characteristics. The structural data clearly indicate some differences in the ligand PEI behavior bound at the two places/regions of glucose oxidase.

Linear and Branched PEIs (Polyethylenimines) and Their Property Space.

A chemical property space defines the adaptability of a molecule to changing conditions and its interaction with other molecular systems determining a pharmacological response. Within a congeneric molecular series (compounds with the same derivatization algorithm and thus the same brute formula) the chemical properties vary in a monotonic manner, i.e., congeneric compounds share the same chemical property space. The chemical property space is a key component in molecular design, where some building blocks are functionalized, i.e., derivatized, and eventually self-assembled in more complex systems, such as enzyme-ligand systems, of which (physico-chemical) properties/bioactivity may be predicted by QSPR/QSAR (quantitative structure-property/activity relationship) studies. The system structure is determined by the binding type (temporal/permanent; electrostatic/covalent) and is reflected in its local electronic (and/or magnetic) properties. Such nano-systems play the role of molecular devices, important in nano-medicine. In the present article, the behavior of polyethylenimine (PEI) macromolecules (linear LPEI and branched BPEI, respectively) with respect to the glucose oxidase enzyme GOx is described in terms of their (interacting) energy, geometry and topology, in an attempt to find the best shape and size of PEIs to be useful for a chosen (nanochemistry) purpose.

QSAR and docking studies of anthraquinone derivatives by similarity cluster prediction.

Forty anthraquinone derivatives have been downloaded from PubChem database and investigated in a quantitative structure-activity relationships (QSAR) study. The models describing log P and LD50 of this set were built up on the hypermolecule scheme that mimics the investigated receptor space; the models were validated by the leave-one-out procedure, in the external test set and in a new version of prediction by using similarity clusters. Molecular docking approach using Lamarckian Genetic Algorithm was made on this class of anthraquinones with respect to 3Q3B receptor. The best scored molecules in the docking assay were used as leaders in the similarity clustering procedure. It is demonstrated that the LD50 data of this set of anthraquinones are related to the binding energies of anthraquinone ligands to the 3Q3B receptor.

Molecular Docking Studies of Flavonoids Derivatives on the Flavonoid 3- O-Glucosyltransferase.

A study of 30 flavonoid derivatives, taken from PubChem database and docked on flavonoid 3-O-glucosyltransferase 3HBF, next submitted to a QSAR study, performed within a hypermolecule frame, to model their LD50 values, is reported. The initial set of molecules was split into a training set and the test set (taken from the best scored molecules in the docking test); the predicted LD50 values, computed on similarity clusters, built up for each of the molecules of the test set, surpassed in accuracy the best model. The binding energies to 3HBF protein, provided by the docking step, are not related to the LD50 of these flavonoids, more protein targets are to be investigated in this respect. However, the docking step was useful in choosing the test set of molecules.

Quantum-mechanical calculations on molecular substructures involved in nanosystems.

In this review article, four ideas are discussed: (a) aromaticity of fullerenes patched with flowers of 6-and 8-membered rings, optimized at the HF and DFT levels of theory, in terms of HOMA and NICS criteria; (b) polybenzene networks, from construction to energetic and vibrational spectra computations; (c) quantum-mechanical calculations on the repeat units of various P-type crystal networks and (d) construction and stability evaluation, at DFTB level of theory, of some exotic allotropes of diamond D5, involved in hyper-graphenes. The overall conclusion was that several of the yet hypothetical molecular nanostructures herein described are serious candidates to the status of real molecules.

QSAR in flavonoids by similarity cluster prediction.

Quantitative Structure-Activity Relationships based on molecular descriptors calculated with correlation weights within the hypermolecule, considered to mimic the investigated correlational space, was performed on a set of 40 flavonoids (PubChem database). The best models describing log P and LD50 of this set of flavonoids were validated by the leave-one-out procedure, in the external test set and in a new version of prediction by using clusters of similar molecules. The best prediction was provided by the similarity cluster procedure.

Hyper-diamonds and dodecahedral architectures by tetrapodal carbon nanotube junctions.

Tetrapodal junctions are used to construct diamond-like networks and dodecahedral architectures. They can be associated with the already synthesized spongy carbon, consisting only of sp2 covalent carbon atoms, and the zeolites, periodic structures in the Euclidean space. In this paper, the structure and stability of two zigzag tetrapodal junctions are discussed. Series of objects are built up by connecting a various number of junctions. Geometry optimization and single point computations (total energy E(tot) and HOMO-LUMO gap energy E(gap)) were performed at the Hartree-Fock level of theory in view of evaluating their stability. The genus of such nanostructures was calculated from the number of consisting tetrapodal junctions.

Molecular design and QSARs/QSPRs with molecular descriptors family.

The aim of the present paper is to present the methodology of the molecular descriptors family (MDF) as an integrative tool in molecular modeling and its abilities as a multivariate QSAR/QSPR modeling tool. An algorithm for extracting useful information from the topological and geometrical representation of chemical compounds was developed and integrated to calculate MDF members. The MDF methodology was implemented and the software is available online (http://l.academicdirect.org/Chemistry/SARs/MDF_SARs/). This integrative tool was developed in order to maximize performance, functionality, efficiency and portability. The MDF methodology is able to provide reliable and valid multiple linear regression models. Furthermore, in many cases, the MDF models were better than the published results in the literature in terms of correlation coefficients (statistically significant Steiger's Z test at a significance level of 5%) and/or in terms of values of information criteria and Kubinyi function. The MDF methodology developed and implemented as a platform for investigating and characterizing quantitative relationships between the chemical structure and the activity/property of active compounds was used on more than 50 study cases. In almost all cases, the methodology allowed obtaining of QSAR/QSPR models improved in explanatory power of structure-activity and structure-property relationships. The algorithms applied in the computation of geometric and topological descriptors (useful in modeling physicochemical or biological properties of molecules) and those used in searching for reliable and valid multiple linear regression models certain enrich the pool of low-cost low-time drug design tools.

Nanotube junctions and the genus of multi-tori.

Carbon nanotube junctions can be modeled by fullerene spanning or by using some operations on map. They can self-assemble into more complex structures, such as finite or infinite high genera multi-tori. Four junctions of tetrahedral and octahedral symmetry, covered by patches consisting only of hexagons, were designed. Their stability is discussed in terms of total energy, evaluated at Hartree-Fock (HF) level of theory, HOMO-LUMO gap, strain energy, HOMA index of aromaticity and the Kekulé structure count. Vibrational spectra of these junctions are given as well. A new multi-toroidal structure, of octahedral symmetry, is presented for the first time. The study on topology of the multi-tori herein designed revealed the relation of these structures with the genus of their embedding surface.

Sumanene Units in P-type Surface Networks.

Sumanene is a synthesized circulene molecule, with formula: 6:(5,6)3. We propose here units for periodic P-type surface networks, based on their stability, evaluated at the Hartree-Fock HF level of theory. Design of the yet hypothetical lattices was performed by using operations on maps, as provided by CVNET and Nano Studio software. The topology of the network was characterized by Omega polynomial.

Deposition of DLC Film on Stainless Steel Substrates Coated by Nickel Using PECVD Method.

Research on diamond-like carbon (DLC) films has been devoted to find both optimized conditions and characteristics of the deposited films on various substrates. In the present work, we investigate the quality of the DLC films grown on stainless steel substrates using different thickness of the nickel nanoparticle layers on the surface. Nickel nanoparticles were sputtered on the stainless steel substrates at 200 °C by a DC-sputtering system to make a good adherence between DLC coating and steel substrates. Atomic Force Microscopy was used to characterize the surface roughness and distribution function of the nickel nanoparticles on the substrate surface. Diamond like carbon films were deposited on stainless steel substrates coated by nickel using pure acetylene and C2H2/H2 with 15% flow ratio by DC-Plasma Enhanced Chemical Vapor Deposition (PECVD) systems. Microstructural analysis by Raman spectroscopy showed a low intensity ratio ID/IG for DLC films by increasing the Ni layer thickness on the stainless steel substrates. Fourier Transforms Infrared spectroscopy (FTIR) evidenced the peaks attributed to C-H bending and stretching vibration modes in the range of 1300-1700 cm-1 and 2700-3100 cm-1, respectively, in good agreement with the Raman spectroscopy and confirmed the DLC growth in all samples.

Polybenzene revisited.

Polybenzene was described by O'Keeffe et al., as an embedding of a 6.82 net in the infinite periodic minimal D-surface, with a single type of carbon atoms and was predicted to have a substantially lower energy per atom in comparison to C60, the reference structure in Nanoscience. They also described a 6.82 net embedded in the periodic minimal P-surface. We give here a rational structure construction for three benzene-based units (a third one described here for the first time in literature) and the corresponding networks. Their stability, relative to C60 but also to diamonds (the classical diamond D6 and the pentagon-based diamond D5), was calculated at the Hartree-Fock level of theory. The results confirmed the previous stability evaluation and support these structures for laboratory preparation. A Graph-theoretical description, in terms of Omega polynomial, of the three infinite networks is also presented.

Counting Polynomials in Tori T(4,4)S[c,n].

A counting polynomial P(x) is a description of a graph property P(G) in terms of a sequence of numbers so that the exponents express the extent of its partitions while the coefficients are related to the number of partitions of a given extent. Basic definitions and some properties are given for two classes of polynomials, called here polynomials of vertex proximity and edge proximity, respectively. Formulas to calculate these polynomials in T(4,4)[c,n] tori are derived by a cutting procedure.

Shell-polynomials and cluj-tehran index in tori t(4,4)s[5,n].

Weighted Hosoya polynomials have been developed by Diudea, in ref. Studia Univ. "Babes-Bolyai", 2002, 47, 131-139. Among various weighting schemes, those polynomials obtained by using Diudea's Shell matrix operator are far more interesting. We present here the Shell-Distance and Shell-Degree-Distance polynomials and close formulas to calculate them and derived Cluj-Tehran CT index in the family of square tiled tori T(4,4)S[5,n]. Applications of the proposed descriptors are also presented.