Title: Palladium catalysts deposited over novel siliceous supports Author: Pavlína Křečková Department: Department of Inorganic Chemistry, Faculty of Science, Charles University in Prague Supervisor: prof. RNDr. Petr Štěpnička, Ph.D. Supervisor's e-mail address: stepnic@natur.cuni.cz Advisor: prof. Ing. Jiří Čejka, DrSc. Advisor's e-mail address: jiri.cejka@jh-inst.cas.cz Abstract Keywords: palladium; deposited catalysts; siliceous supports; C-C coupling reaction. The aim of this work was to prepare a deposited palladium catalyst from Pd2+ and twodimensional zeolite ITQ-2 bearing ≡SiCH2CH2CH2NHCH2CH2NEt2 groups at the surface. This catalys has been evaluated in the Heck reaction between n-butyl-acrylate and brombenzene to give n-butyl-cinnamate. Influence of the reaction temperature, the solvent, base and hydrate water on the progress of the Heck reaction was investigated. This work also deals with the possibility of recycling the catalyst and, due to its nature, with the quantity of leached palladium into the reaction mixture. Furthermore, the work describes differences in morphology between the fresh and a recycled catalyst.

Compounds with carboxylic and amidic functions belong to basic structural blocks, which are used for construction of functional molecules in organic, organometallic and also in carborane chemistry. However, considering cobalt bis(dicarbollide)(1-) ion, the synthetic ways to these derivatives have been virtually unknown. A published procedure on lithiation in THF and reaction with CO2 leading to mono- and dicarboxylic acids had failed in our hands. Nevertheless, a detailed revision of the experimental conditions provided finally good yields of mixture of both acids, which could be separated by chromatography and crystallization, and compound of general formulation [(1-HOOC-1,2-C2B9H10)(1',2'-C2B9H10)-3,3'-Co(III)]- and stereoisomeric mixture of [(HOOC)2-(1,2-C2B9H10)2-3,3'-Co(III)]- were characterized for the first time by combination of NMR, MS and HPLC. Also, the carboxylic acid derivatives with methylene and ethylene connectors of the general formula [(1-HOOC-(CH2)n-1,2- C2B9H10)(1',2'-C2B9H10)-3,3'-Co(III)]- were prepared by lithiation of Cs1 in DME at low temperatures followed by reaction with BrCH2COOEt and subsequent hydrolysis of the resulting ester or by oxidation of the respective propylhydroxy derivative. The acids were converted to reactive p-nitrophenyl esters...

β-N-acetylhexosaminidases (EC 3.2.1.52, GH 20) are retaining exo-glycosidases that in vivo cleavage both β-N-acetylglucosamine (GlcNAc) or β-N-acetylgalactosamine (GalNAc) residues fom glycostructures. Under suitable reaction conditions, these enzymes are able to synthesize the glycosidic bond in good yields. Substitution of selected amino acid(s) in the emzyme active site by site-directed mutagenesis may change the enzyme's substrate specificity or suppress the hydrolytic activity of the enzyme in favor of synthesis. The present thesis deals with three mutant β-N-acetylhexosaminidases from Talaromyces flavus, in which the amino acid residues responsible for binding to C-4 hydroxyl of the substrate (Arg218, Glu546) were exchanged for amino acids proposed on the basis of molecular modeling. The effect of introduced single point mutations on substrate specificity of prepared enzymes was studied. Mutant β-N-acetylhexosaminidases were heterologously expressed in Pichia pastoris and characterized. Furthermore, transglycosylation reactions with these enzymes were performed. The prepared carbohydrate products were characterized by NMR.

While recent neural sequence-to-sequence models have greatly improved the quality of speech synthesis, there has not been a system capable of fast training, fast inference and high-quality audio synthesis at the same time. In this the- sis, we present a neural speech synthesis system capable of high-quality faster- than-real-time spectrogram synthesis, with low requirements on computational resources and fast training time. Our system consists of a teacher and a student network. The teacher model is used to extract alignment between the text to synthesize and the corresponding spectrogram. The student uses the alignments from the teacher model to synthesize mel-scale spectrograms from a phonemic representation of the input text efficiently. Both systems utilize simple convo- lutional layers. We train both systems on the english LJSpeech dataset. The quality of samples synthesized by our model was rated significantly higher than baseline models. Our model can be efficiently trained on a single GPU and can run in real time even on a CPU. 1

In this diploma Thesis I dealt with the synthesis of nitrogen compounds that contain quaternary carbon centres and their structure resembles natural substances, specifically alkaloids from the Amaryllidaceae family. Halocarbocyclization or Heck reaction was planned as a key step of the synthesis, which should lead to the formation of quaternary carbon centres. First, starting nitriles and esters with five- and six-membered rings were prepared. Subsequently, the method for the preparation of a stable bicyclic ketone with a five-membered ring from an ester was optimized. The next part of the Thesis is devoted to the synthesis of a substrate for the Heck reaction from the prepared ketone, which was then used for the preparation of the alkaloid skeleton with a quaternary carbon centre, unfortunately without success so far. In the last part of the work, a bicyclic ketone with a six-membered ring, from which it is possible to prepare a substrate for the Heck reaction in several steps already used for five-membered substances, was prepared from the nitrile by carbopalladation. Key words: synthesis, quaternary carbon centres, Heck reaction, halocarbocyclisation, Amaryllidaceae alkaloids

The diploma thesis deals with the development of a new method of synthetic peptide separation using electrostatic binding. It deals with the development of a multi-charged anchored linker attachable to a synthetic peptide after its synthesis on solid phase. Subsequently, it investigates the separation using electrostatic binding; however, after peptide cleavage from the anchored linker and Boc deprotection with TFA, the peptide formed so strong electrostatic interactions with the cation exchanger that we were unable to recover it. The thesis is finished with discussion and method improvement proposal. Key words: peptide separation, solid-phase peptide synthesis, SPPS, electrostatic binding

DNA-encoded peptide libraries are the basis for in vitro selection methods that use various biological systems (phage display; yeast display; mRNA display). Despite the great success of these selection methods, their obvious disadvantage is the limited number of building blocks, which consist of only twenty proteinogenic amino acids. The involvement of other non- proteinogenic amino acids and other building blocks could significantly expand the range of possible applications of these selection methods. For example, the introduction of chemical modifications in amino acid side chains in such libraries would allow the effective study of post-translational modifications (phosphorylation, acylation, glycosylation, methylation, etc.) in living organisms. The aim of this work was to develop a method for preparation of a fully synthetic DNA encoded library of peptides. The basic steps for the preparation were the chemical synthesis of the peptide and associated enzymatic synthesis of encoding DNA. Compatibility of chemical reactions with DNA is essential for the synthesis of DNA-encoded peptide libraries. Because the final acidic deprotection of the side chains in the peptide is not compatible with DNA, two approaches have been tested to overcome this problem. The first was an attempt to develop finer...

This thesis deals with the synthesis of polycyclic compounds using tandem cyclisation/Suzuki cross-coupling, by modification of the method developed by our research group. This thesis mainly investigates the scope of the tandem reaction with aliphatic boronic acids or their esters. The products of the palladium catalyzed reaction then underwent electrocyclization to form an aromatic ring. The polycyclic skeleton, containing the aromatic core, is found in a considerable amount of natural substances with biological activity. The structure of the compounds prepared herein is similar to miltiorin D or commiphorane A.

This thesis deals with single-electron transfer (SET) radical processes mediated by ferrocenium hexafluorophosphate and TEMPO and their application in the total synthesis of natural products. Asymmetric aminooxygenation methodology for the synthesis of anti-β-amino-α-hydroxy acid derivatives has been developed by utilizing a highly diastereoselective aza-Michael addition of chiral lithium amides to various α,β-unsaturated esters or amides/SET oxidation/radical α-oxygenation. The potential of this methodology was demonstrated in short total syntheses of the anti-β-amino-α-hydroxy acid fragments of the macrocyclic (depsi)peptides perthamide C and largamide H, and (-)-cytoxazone, which is a selective modulator of TH2 cytokine secretion. The SET-catalyzed asymmetric tandem lithium amide conjugate addition/5-exo radical cyclization/oxygenation reactions were applied in the synthesis of highly substituted pyrrolidines, azabicyclo[n.3.0]alkanes and spiropyrrolidines. An enantioselective total synthesis of the pyrrolidine alkaloid (-)-α-kainic acid was accomplished by employing the SET-catalyzed 5-exo radical cyclization/oxygenation.

This work explores multilingual speech synthesis. We compare three models based on Tacotron that utilize various levels of parameter sharing. Two of them follow recent multilingual text-to-speech systems. The first one makes use of a fully-shared encoder and an adversarial classifier that removes speaker-dependent information from the encoder. The other uses language-specific encoders. We introduce a new approach that combines the best of both previous methods. It enables effective parameter sharing using a meta- learning technique, preserves encoder's flexibility, and actively removes speaker-specific information in the encoder. We compare the three models on two tasks. The first one aims at joint multilingual training on ten languages and reveals their knowledge-sharing abilities. The second concerns code-switching. We show that our model effectively shares information across languages, and according to a subjective evaluation test, it produces more natural and accurate code-switching speech.