Stepan Sklenak

Application of recurrent neural networks in chemistry. Prediction and classification of carbon-13 NMR chemical shifts in a series of monosubstituted benzenes

The recurrent neural network is a feed-forward network ascribed to a parent neural network with feed-back connections (or in another term, oriented cycles). Its adaptation is performed by an analog of the standard back-propagation adaptation method. The recurrent neural network approach is illustrated by prediction and classification of 13C NMR chemical shifts in a series of monosubstituted benzenes. The descriptors (input activities) of functional groups are determined by 11 nonnegative integers that correspond to numbers of appearance of some substructural features in the corresponding molecular graphs. The obtained results indicate that these descriptors properly describe the basic physical and chemical nature of functional groups.

Ab initio study of small organic azides

Abstract Ab initio MP2(full)/TZ2P//MP2(full)/TZ2P calculations of the following organic alkyl azides were carried out: hydrazoic acid (1), methyl azide (2), ethyl azide (3), 3-azidopropine (4), azidoacetonitrile (5), azido-2-butine (6), allylazide (7), 2-azido-1, 3-butadiene (8). Fully optimized geometries of all rotational conformers of these molecules were calculated. Ab initio MP2 analytical force fields were computed and the theoretical vibrational wavenumbers were obtained. Good agreement is found between the theoretical and experimental IR/Raman and microwave spectra, and also for structures obtained by electron diffraction.

Polycyclic Arene Episulfides. Attempted Synthesis, Molecular Orbital Calculations and Comparison with Arene Oxides

In pursuit of the elusive polycyclic arene episulfides, the synthesis of phenanthrene-9,10-episulfide (5b) was attempted. However, the reactions of phenanthrene-9,10-oxide and of phenanthrene with sulfur transfer agents and of 5,7-dihydrodibenzo[c,e]thiepine (9) with butyllithium did not afford 5b, although its intermediacy could be detected by NMR and inferred from the products isolated. Quantum-mechanical ab initio and density functional methods were used to calculate the thermodynamic stability of representative arene episulfides and arene oxides. The arene episulfides were found to be thermodynamically significantly less stable than the corresponding oxides, towards elimination of sulfur and oxygen respectively. K-region arene episulfides are found to be thermodynamically more stable towards sulfur extrusion than their bay-region isomers. The thermodynamic stabilities of analogous arene oxides and arene episulfides parallel each other, and the relative thermodynamic stabilities of arene oxides and episulfides can be deduced from the degree of the aromatic character of the rings which do not carry the heteroatom.

Application of neural networks with feedback connections in chemistry: prediction of 13C NMR chemical shifts in a series of monosubstituted benzenes

Abstract A simple steepest-descent adaptation process of neural networks with feedback connections (i.e. with oriented cycles) is described. The method is illustrated with an application of this type of neural network to the prediction of 13C NMR chemical shifts in a series of monosubstituted benzenes.

Location of Framework Al Atoms in the Channels of ZSM‐5: Effect of the (Hydrothermal) Synthesis

(27) Al 3Q MAS NMR and UV/Vis spectroscopy with bare Co(II) ions as probes of Al pairs in the zeolite framework were employed to analyze the location of framework Al atoms in the channel system of zeolite ZSM-5. Furthermore, the effect of Na(+) ions together with tetrapropylammonium cation (TPA(+)) in the ZSM-5 synthesis gel on the location of Al in the channel system was investigated. Zeolites prepared using exclusively TPA(+) as a structure-directing agent (i.e., in the absence of Na(+) ions) led to 55-90% of Al atoms located at the channel intersection, regardless the presence or absence of Al pairs [Al-O-(Si-O)2 -Al sequences in one ring] in the zeolite framework. The presence of Na(+) ions in the synthesis gel did not modify the Al location at the channel intersection (55-95% of Al atoms) and led only to changes in i) the distribution of framework Al atoms between Al pairs (decrease) and single isolated Al atoms (increase), and ii) the siting of Al in distinguishable framework tetrahedral sites.

Role of Glutamate 64 in the Activation of the Prodrug 5-fluorocytosine by Yeast Cytosine Deaminase

Yeast cytosine deaminase (yCD) catalyzes the hydrolytic deamination of cytosine to uracil as well as the deamination of the prodrug 5-fluorocytosine (5FC) to the anticancer drug 5-fluorouracil. In this study, the role of Glu64 in the activation of the prodrug 5FC was investigated by site-directed mutagenesis, biochemical, nuclear magnetic resonance (NMR), and computational studies. Steady-state kinetics studies showed that the mutation of Glu64 causes a dramatic decrease in k(cat) and a dramatic increase in K(m), indicating Glu64 is important for both binding and catalysis in the activation of 5FC. (19)F NMR experiments showed that binding of the inhibitor 5-fluoro-1H-pyrimidin-2-one (5FPy) to the wild-type yCD causes an upfield shift, indicating that the bound inhibitor is in the hydrated form, mimicking the transition state or the tetrahedral intermediate in the activation of 5FC. However, binding of 5FPy to the E64A mutant enzyme causes a downfield shift, indicating that the bound 5FPy remains in an unhydrated form in the complex with the mutant enzyme. (1)H and (15)N NMR analysis revealed trans-hydrogen bond D/H isotope effects on the hydrogen of the amide of Glu64, indicating that the carboxylate of Glu64 forms two hydrogen bonds with the hydrated 5FPy. ONIOM calculations showed that the wild-type yCD complex with the hydrated form of the inhibitor 1H-pyrimidin-2-one is more stable than the initial binding complex, and in contrast, with the E64A mutant enzyme, the hydrated inhibitor is no longer favored and the conversion has a higher activation energy, as well. The hydrated inhibitor is stabilized in the wild-type yCD by two hydrogen bonds between it and the carboxylate of Glu64 as revealed by (1)H and (15)N NMR analysis. To explore the functional role of Glu64 in catalysis, we investigated the deamination of cytosine catalyzed by the E64A mutant by ONIOM calculations. The results showed that without the assistance of Glu64, both proton transfers before and after the formation of the tetrahedral reaction intermediate become partially rate-limiting steps. The results of the experimental and computational studies together indicate that Glu64 plays a critical role in both the binding and the chemical transformation in the conversion of the prodrug 5FC to the anticancer drug 5-fluorouracil.

Mechanism of framework oxygen exchange in Fe-zeolites: a combined DFT and mass spectrometry study.

The role of framework oxygen atoms in N(2)O decomposition [N(2)O(g)→N(2)(g) and 1/2O(2)(g)] over Fe-ferrierite is investigated employing a combined experimental (N(2)(18)O decomposition in batch experiments followed by mass spectroscopy measurements) and theoretical (density functional theory calculations) approach. The occurrence of the isotope exchange indicates that framework oxygen atoms are involved in the N(2)O decomposition catalyzed by Fe-ferrierite. Our study, using an Fe-ferrierite sample with iron exclusively present as Fe(II) cations accommodated in the cationic sites, shows that the mobility of framework oxygen atoms in the temperature range: 553 to 593 K is limited to the four framework oxygen atoms of the two AlO(4)(-) tetrahedra forming cationic sites that accomodate Fe(II). They exchange with the Fe extra-framework (18)O atom originating from the decomposed N(2)(18)O. We found, using DFT calculations, that O(2) molecules facilitate the oxygen exchange. However, the corresponding calculated energy barrier of 87 kcal  mol(-1) is still very high and it is higher than the assumed experimental value based on the occurrence of the sluggish oxygen exchange at 553 K.

TNU‐9 Zeolite: Aluminum Distribution and Extra‐Framework Sites of Divalent Cations

The TNU-9 zeolite (TUN framework) is one of the most complex zeolites known. It represents a highly promising matrix for both acid and redox catalytic reactions. We present here a newly developed approach involving the use of 29 Si and 27 Al (3Q) MAS NMR spectroscopy, CoII as probes monitored by UV/Vis and FTIR spectroscopy, and extensive periodic DFT calculations, including molecular dynamics, to investigating the aluminum distribution in the TUN framework and the location of aluminum pairs and divalent cations in extra-framework cationic positions. Our study reveals that 40 and 60 % of aluminum atoms in the TNU-9 zeolite are isolated single aluminum atoms and aluminum pairs, respectively. The aluminum pairs are present in two types of six-membered rings forming the corresponding α and β (15 and 85 %, respectively, of aluminum pairs) sites of bare divalent cations. The α site is located on the TUN straight channel wall and it connects two channel intersections. The suggested near-planar β site is present at the channel intersection.

Aluminum siting in the framework of silicon rich zeolites. A ZSM-5 study

Abstract 27 Al 3Q MAS NMR spectroscopy and DFT/MM calculations were employed to study the Al siting in eleven differently synthesized samples of the ZSM-5 zeolite. The results of our study reveal that the occupation of the distinguishable framework T sites by Al and the concentration of Al in these T sites are neither random nor controlled by a simple rule. They both depend on the conditions of the zeolite synthesis. We further found that at least 10 out of the 24 distinguishable framework T sites are occupied by Al in our ZSM-5 samples.

CuNO2 and Cu(+)NO2 Revisited: A Comparative ab Initio and DFT Study.

We have reinvestigated CuNO2 and Cu(+)NO2 at ab initio as well as at pure and hybrid DFT levels of approximation employing large ANO basis sets. The systems were fully optimized using the CCSD(T), QCISD(T), BPW91, PBE, PBE0, and B3LYP methods. Several stationary points (minima and transition structures) were found on the related potential energy surfaces (PES). The C2v bidentate η(2)-O,O isomer is calculated to be the most stable species on the CuNO2 PES, followed by two monodentate isomers [Formula: see text] the Cs η(1)-O and C2v η(1)-N species which are higher in energy by 12 and 14 kcal/mol, respectively, at CCSD(T)/Basis-II (where Basis-II is 21s15p10d6f4g/8s7p5d3f2g for Cu; 14s9p4d3f/5s4p3d2f for O and N). On the Cu(+)NO2 PES, the Cs monodentate η(1)-O trans (0 kcal/mol) and cis (+3 kcal/mol at CCSD(T)/Basis-II) isomers are found, followed by the C2v monodentate η(1)-N isomer (+14 kcal/mol at CCSD/Basis-II). In contrast to the pure DFT, the hybrid DFT methods perform reasonably well for predicting the relative stabilities (except for η(1)-N of CuNO2) and structures; however, their predictions of the bond dissociation energies are less reliable (for CuNO2 the difference is as much as 10 kcal/mol compared to the CCSD(T) values). The performance of the QCISD(T) method was analyzed, and, furthermore, the issue of symmetry breaking was investigated.