Search

We describe the synthesis together with the results of anti-HIV-1 activity and gp120-monolayer binding experiments of new galactosyl amphiphiles, analogues of galactosylceramide, an alternative receptor used by HIV to infect CD4 negative cells. These compounds consist of single- and double-chain amphiphiles containing one or two galactose residues. To favor their clustering into galactosyl-rich microdomains, their molecular structure contains also an amino group or several hydroxyls or anionic groups, such as carboxylate, sulfate, sulfonate, and phosphate. Among the 12 new galactosylated compounds reported, a specific anti-HIV activity, although moderate (IC50 from 10 to 50 μM), was detected only for three of them, i.e., I-GalSer[CO2Na][C14], II-GalSer[C14][C7SO3Na], and II-GalSer[C2SO4Na][C14], which contain an anionic group. The marked increase of surface pressure which was observed upon addition of gp120 into the aqueous subphase underneath the monolayers containing these galactolipids indicated gp120 insertion into the monolayers, suggesting that binding of these three derivatives to HIV-1 gp120 may be responsible for their anti-HIV activity.

There is an abundance of structural molecular descriptors of various forms that have been proposed and tested over the years. Very often different descriptors represent, more or less, the same aspects of molecular structures and, thus, they have diminished discriminating power for the identification of different structural features that might contribute to the molecular property, or activity of interest. Therefore, it is essential that noncorrelated descriptors be employed to ensure the wider and the less inflated possible coverage of the chemical space. The most usual approach for reducing the number of descriptors and employing noncorrelated (or orthogonal) descriptors involves principal component analysis (PCA) or other factor analytical techniques. In this work we present an approach for determining relationships (groupings) among 240 graph-theoretical descriptors, as a means for selecting nonredundant ones, based on the application of cluster analysis (CA). To remove inherent biases and particularities of different CA algorithms, several clustering solutions, using these algorithms, were “hybridized” to obtain a reliable and confident overall solution concerning how the interrelationships within the data are structured. The calculated correlation coefficients between descriptors were used as a reference for a discussion on the different CA methods employed, and the resulted clusters of descriptors were statistically analyzed for deriving the intercorrelations between the different operators, weighting schemes and matrices used for the computation of these descriptors.

During the search for new structural descriptors we have defined the information-theory operators U(M), V(M), X(M), and Y(M), that are computed from atomic invariants and measure the information content of the elements of molecular matrices. Structural descriptors computed with these four information-theory operators are used to develop structure−property models for the boiling temperature, molar heat capacity, standard Gibbs energy of formation, vaporization enthalpy, refractive index, and density of alkanes. The information-theory operators were applied to six molecular matrices, namely, the distance D, the reciprocal distance RD, the distance-path Dp, the reciprocal distance-path RDp, the path Szeged Szp, and the reciprocal path Szeged RSzp matrices. In combination with other topological indices, the information-theory indices offer good structure−property models for all six alkane properties investigated in this study.

Organic compounds containing heteroatoms or multiple bonds can be conveniently represented as vertex- and edge-weighted molecular graphs. These atom and bond parameters can be computed for any organic compound with two parameter sets that we have recently defined, namely, the relative electronegativity X and the relative covalent radius Y weighting schemes. Structural descriptors computed with these two weighting schemes and the previously defined atomic number Z parameter set are used to develop quantitative structure−retention relationship (QSRR) models for alkylphenols in gas−liquid chromatography. The QSRR models are generated with structural descriptors computed with several newly introduced graph operators, namely, the Wiener, hyper-Wiener, minimum eigenvalue, maximum eigenvalue, Ivanciuc−Balaban, and information on distance operators. These molecular graph operators were applied to the distance D and the reciprocal distance RD matrixes.

This paper presents a new methodology of chemical substructure recognition by interpretation of an infrared spectrum. The approach in spectrum interpretation is based on the determination of functional groups, which may be present or absent in compounds whose structure is unknown. The process of searching for spectrum−substructure correlation is realized by application of a statistical algorithm. In this method, correlations are generalized and condensed into a set of interpretation rules which are applied to the interpretation of an unknown compound's spectrum in order to predict whether the respective substructures are present or absent in the unknown molecule.

In the pharmaceutical industry, the virtual screening of combinatorial libraries is used to rationally select compounds for biological testing from databases of hundreds of thousands of compounds. In addition to structural descriptors, such as fingerprints and pharmacophores, the application of relatively simple structural descriptors traditionally used in quantitative structure−activity studies offers speed and efficiency for rapidly measuring the molecular diversity of such collections. We explore new topological indices computed from the molecular graph as potential structural descriptors for the characterization of molecular diversity. A database of 2000 compounds randomly selected from the National Cancer Institute AIDS database was used to measure the intercorrelation of the descriptors. The initial collection of 240 structural descriptors was reduced to several quasi-orthogonal sets of up to 9 descriptors, using different thresholds for the maximum intercorrelation coefficient.

The regulatory protein Tat is essential for viral gene expression and replication of the human immunodeficiency virus type 1 (HIV-1). Tat transactivates the HIV-1 long terminal repeat (LTR) via its binding to the transactivation responsive element (TAR) and increases the viral transcription. Studies have shown that the binding of arginine and arginine derivatives induces a conformational change of the TAR RNA at the Tat-binding site. The unpaired A17 residue delimits a small cavity which constitutes a receptor site for small molecules, especially for ethidium bromide. These binding characteristics have prompted us to design a series of ethidium−arginine conjugates capable of interacting with the TAR RNA. Here we report the synthesis of six ethidium derivatives equipped with arginine side chains. These molecules were biologically evaluated, and two compounds (17 and 20) exhibited in vitro anti-HIV-1 activity at micromolar concentration, without toxicity (up to 100 μM concentration). Melting temperature studies indicated that the most active molecule (20) bound strongly to TAR in vitro. RNase protection experiments agreed with the molecular modeling studies which suggested that the ethidium moiety of 20 was inserted next to the A17 residue while the arginine side chain occupied the pyrimidine bulge.

A QSPR (quantitative structure−property relationship) model for the estimation of the liquid viscosity of a large variety of organic compounds was established using the CODESSA (comprehensive descriptors for structural and statistical analysis) approach. The final model was developed with a calibration set containing 337 compounds. The multilinear regression (MLR) equation that relates ln η to five molecular descriptors has s = 0.37 and r = 0.920. The five theoretical parameters used in the QSPR model are molecular weight, Randić connectivity index of order 3, hydrogen-donor charged surface area HDCA-2 (electrostatic), maximum electrophilic reactivity index for a carbon atom, and maximum atomic orbital electronic population. The predictive ability of the MLR model was tested by the leave-20%-out cross-validation method, showing that the QSPR model is stable and can be used to obtain good predictions for compounds that were not used in the model calibration. The cross-validation statistical indices show a small decrease when compared with those obtained in the calibration phase (scv = 0.38, rcv = 0.917). The QSPR models developed with CODESSA allow accurate computation of the liquid viscosity of organic compounds using simple constitutional descriptors and quantum indices which can be computed with standard quantum chemistry packages.

The 13C NMR chemical shift of sp2carbon atoms in acyclic alkenes was estimated with multilayer feedforward artificial neural networks (ANNs) and multilinear regression (MLR), using as structural descriptors a vector made of 12 components encoding the environment of the resonating carbon atom. The neural network quantitative model provides better results than the MLR model calibrated with the same data. The predictive ability of both the ANN and MLR models was tested by the leave-20%-out (L20%O) cross-validation method, demonstrating the superior performance of the neural model. The number of neurons in the hidden layer was varied between 2 and 7, and three activation functions were tested in the neural model:  the hyperbolic tangent or a bell-shaped function for the hidden layer and a linear or a hyperbolic tangent function for the output layer. All four combinations of activation functions give close results in the calibration of the ANN model, while for the prediction a linear output function performs better than a hyperbolic tangent one, but from a statistical point of view one could not choose a particular combination against the others. For the ANNs with four neurons in the hidden layer, the standard deviation for calibration ranges between 0.59 and 0.63 ppm, while for prediction it lies between 0.89 and 1.07 ppm. We propose a parallel use of the four ANNs for the prediction of unknown shifts, because the mean of the four predictions exhibit a smaller number of outliers with lower residuals. The present model is compared with three additive schemes for the calculation of the sp213C NMR chemical shifts, and the statistical analysis of the results demonstrates that the ANN model gives better predictions than the classical ones.

Two new computer methods designed to reveal structural features of unknown compounds by means of low resolution mass spectra are presented. Both methods use the results of a spectral similarity search in a mass spectral database. The first one proceeds by intersecting selected structures in order to find maximal common substructures, while the second proceeds by decomposing these structures to derive fragments following a model of primary fragmentation of organic molecules. Reliability of the revealed fragments is estimated by comparing an unknown compound's spectrum with the computed spectral images of each fragment. The usefulness and limitations of the two proposed methods are estimated by using a set of test examples. In many cases the two methods are complementary, whereas overall, the second looks more promising both for revealing large structural fragments and for generation of candidate structures, because the fragments revealed have only one or two free valences and rarely overlap one another.

The degeneracy of topological distance sum (DS) and distance degree sequence (DDS) of small fullerenes is analyzed. The degeneracy of the two topological descriptors is determined by comparing the topological equivalence classes of the fullerene atoms with the DS and DDS classes, respectively. On the basis of DS and DDS degenerated fullerene graphs and by using two previously demonstrated theorems it is possible to generate pairs of graphs with degenerated topological indices based on DS (e.g., W, J) and on DDS (e.g., U, V, X, Y).

The ¹³C NMR chemical shift of sp³ carbon atoms situated in the α position relative to the double bond in acyclic alkenes was estimated with multilayer feedforward artificial neural networks (ANNs) and multilinear regression (MLR), using as structural descriptors a topo-stereochemical code which characterizes the environment of the resonating carbon atom. The predictive ability of the two models was tested by the leave-20%-out cross-validation method. The neural model provides better results than the MLR model both in calibration and in cross-validation, demonstrating that there exists a nonlinear relationship between the structural descriptors and the investigated ¹³C NMR chemical shift and that the neural model is capable to capture such a relationship in a simple and effective way. A comparison between a general model for the estimation of the ¹³C NMR chemical shift and the ANN model indicates that general models are outperformed by more specific models, and in order to improve the predictions a possible way is to develop environment-specific models. The approach proposed in this paper can be used in automated spectra interpretation or computer-assisted structure elucidation to constrain the number of possible candidates generated from the experimental spectra.

An artificial neural network (ANN) has been trained to recognize the presence or absence of specific structural features (SF) in trimethylsilyl derivatives of organic acids from their mass spectra. The input vector is constructed without knowledge of the molecular ion, which is generally not observed in the spectra of these compounds. The results are used in conjunction with a classical search in a spectral library to identify organic acids in biological fluids for rapid acidemias diagnosis.

A new general molecular coding Prolog program is used in theoretical studies of the properties of fullerene isomers and derivatives to identify the topological equivalence classes of atoms and bonds in C20−C60 fullerenes. The symmetry perception algorithm is based on an exhaustive search of a minimal code. During this search all canonical labeled mapping trees are generated, allowing construction of equivalence classes of atoms and bonds. Although the present algorithm makes a complete search of canonical labelings in the molecular graph, computational time is maintained within reasonable limits even for large regular graphs such as fullerenes. In the case of fullerenes, topological equivalence classes of atoms allow an easy generation of heterofullerene isomers, while the bond classes allow for the generation of addition or cycloaddition isomers. Also, the number of atom classes and the number of atoms in each class gives important information on the number of ¹³C NMR signals and intensity patterns in fullerenes.

The ¹³C NMR chemical shift of sp² carbon atoms in acyclic alkenes was estimated with multilayer feedforward artificial neural networks (ANNs) and multilinear regression (MLR), using as structural descriptors a vector made of 12 components encoding the environment of the resonating carbon atom. The neural network quantitative model provides better results than the MLR model calibrated with the same data. The predictive ability of both the ANN and MLR models was tested by the leave-20%-out (L20%O) cross-validation method, demonstrating the superior performance of the neural model. The number of neurons in the hidden layer was varied between 2 and 7, and three activation functions were tested in the neural model:  the hyperbolic tangent or a bell-shaped function for the hidden layer and a linear or a hyperbolic tangent function for the output layer. All four combinations of activation functions give close results in the calibration of the ANN model, while for the prediction a linear output function performs better than a hyperbolic tangent one, but from a statistical point of view one could not choose a particular combination against the others. For the ANNs with four neurons in the hidden layer, the standard deviation for calibration ranges between 0.59 and 0.63 ppm, while for prediction it lies between 0.89 and 1.07 ppm. We propose a parallel use of the four ANNs for the prediction of unknown shifts, because the mean of the four predictions exhibit a smaller number of outliers with lower residuals. The present model is compared with three additive schemes for the calculation of the sp^{2 13}C NMR chemical shifts, and the statistical analysis of the results demonstrates that the ANN model gives better predictions than the classical ones.

The 13C NMR chemical shift of sp3carbon atoms situated in the position relative to the double bond in acyclic alkenes was estimated with multilayer feedforward artificial neural networks (ANNs) and multilinear regression (MLR), using as structural descriptors a topo-stereochemical code which characterizes the environment of the resonating carbon atom. The predictive ability of the two models was tested by the leave-20%-out cross-validation method. The neural model provides better results than the MLR model both in calibration and in cross-validation, demonstrating that there exists a nonlinear relationship between the structural descriptors and the investigated 13C NMR chemical shift and that the neural model is capable to capture such a relationship in a simple and effective way. A comparison between a general model for the estimation of the 13C NMR chemical shift and the ANN model indicates that general models are outperformed by more specific models, and in order to improve the predictions a possible way is to develop environment-specific models. The approach proposed in this paper can be used in automated spectra interpretation or computer-assisted structure elucidation to constrain the number of possible candidates generated from the experimental spectra.