|
Studies of NK cell receptors and other proteins using recombinant expression and mass spectrometry
Kavan, Daniel ; Bezouška, Karel (advisor) ; Tučková, Ludmila (referee) ; Řehulka, Pavel (referee)
Charles University in Prague Faculty of Science Department of Biochemistry Studies of NK cell receptors and other proteins using recombinant expressions and mass spectrometry Summary of Ph. D. Thesis Daniel Kavan Supervisor: Prof. RNDr. Karel Bezouška, DSc. Prague 2010 Daniel Kavan Introduction Introduction NK cells and CD69 as one of their surface receptors Natural killer cells (NK cells) are the subpopulation of large granular lymfocytes, which lacks the surface receptors typical for B cells or T cells. They are characterized by the presence of NKp46 and NKp30 [Moretta L. et. al. 2002], however. They were named natural killers according to their function in the organism, as they do not need any activation and nevertheless they are able to eliminate abnormal (i. e. infected or transformed) cells from the tissue [Kiessling R. et. al. 1975]. This function is dependent on scanning the major histocompatibility complex (MHC) class I molecules of ambient cells. The resulting action (killing or not killing the target cell) is dependent on the balance of activating and inhibiting signals mediated by the NK cell surface receptors and forwarded to the specific signaling pathway [Raulet D. H. et. al. 2001]. Specificity of NK cells is not based only on one type of antigen receptor as it is in case of T and B cells,...
|
|
Programmes of monitoring of bird occurrence and abundance as a source of ecological data
Mourková, Jindra ; Fuchs, Roman (advisor) ; Reif, Jiří (referee) ; Bejček, Vladimír (referee)
5 regionálních atlasech (Flousek & Gramsz 1999, Hromádko et al. 2005), umožňuje lépe zachytit proměnlivost výskytu a početnosti ptáků, která je daná mozaikovitostí studovaného území. Kromě údajů o prokázanosti hnízdění v jednotlivých kvadrátech síťových map přináší také odhad početnosti v celém území CHKO, charakteristiku výskytu a prostředí využívaného druhem na daném území a údaje o hnízdních hustotách v různých typech prostředí. Kvantitativní mapy pro jednotlivé druhy nejsou bohužel dosud zpracovány. SUMMARY OF THE THESIS In the thesis, regional data on bird communities were studied in terms of species richness and occurrence of bird species in dependence on habitat and, in case of wintering birds, also on climatic conditions. The regional grid atlas of Prague (Fuchs et al. 2002), records from Wintering waterbird census in Central Bohemia, own data on the breeding population of the Mute Swan in the Třeboň region and data from individual marking of Mute Swans provided by the Bird Ringing Centre of the National Museum were used as the sources of data on the occurrence and abundance of bird species. The thesis also includes a part of the text of the prepared Atlas of breeding distribution of birds in the Třeboň region. The first contribution (Mourková J., Fuchs R., Sedláček O., Janovský Z., Exnerová A.,...
|
|
Vodomilovití brouci (Coleoptera: Hydrophiloidea): taxomomie, fylogeneze a fosilní záznam vybraných skupin
Fikáček, Martin ; Prokop, Jakub (advisor) ; Engel, Michael (referee) ; Balke, Michael (referee)
The presented PhD thesis consists of two parts. The first part concerns the fos- sil record of the hydrophiloid beetles. Its introductory chapter summarizes the current author's knowledge of the topic, based on the review of all available literature and his published and unpublished data. Historical review is provided, problems concerning the study of the hydrophiloid fossil taxa and the relevance of the fossil record are discussed. Two published papers and two manuscripts concerning the systematics and taxonomy of selected hydrophilid beetles from European Cenozoic deposits are attached. As a part of these results, the diver- sification of the greater hydrophilines lineage of the subtribe Hydrophilina is dated back to the middle Eocene and the occurrence of the currently tropical ge- nus Hydrobiomorpha in Europe is proved from Eocene to the latest Oligocene. The first fossil of the genus Limnoxenus (subtribe Hydrobiusina) is recorded from the late Oligocene locality of Aix-en-Provence, its taxonomic placement is tested by the phylogenetic analysis combining modern and fossil taxa and its relevance is discussed. The second part concerns selected taxa of the terrestrial hydrophilids of the subfamily Sphaeridiinae. Its introductory chapter briefly summarizes the...
|
|
Chipped stone industry in the context of the Early Bronze Age
Rychtaříková, Tereza ; Limburský, Petr (advisor) ; Šebela, Lubomír (referee)
This diploma thesis focuses on studies of chipped stone industry of the Early Bronze Age in Bohemia and deals with questions connected to its occurrence in settlement and burial sites. The first part of this study presents the results of artefact analysis. The anylised complex of artefacts consists of 108 items and originates in following localities: Kolín - ringroad, Pátek near Poděbrady, Polepy near Kolín, Velké Přílepy a Vliněves. The analysis included the typological and morphological characteristics, used material and the links to their position in settlement objects and other finds. The second part of this diploma thesis summarizes present knowledge of prestigious artefacts such as silicite daggers, axes and arrowheads. Furthermore, it investigates the matter of contacts between Central Bohemia and Moravia on the basis of Moravian stone materials distribution. A catalogue of chipped stone industy from Central Bohemia is also included in this diploma thesis.
|
|
Generation of protein sequences with a given characteristic
Hrbáň, Hugo ; Hoksza, David (advisor) ; Lokoč, Jakub (referee)
Proteins are essential for life as they play a fundamental role in many biological processes. Designing novel proteins with a desired function is an important problem in drug development and biological research. Large databases of protein sequences can be used to train large language models adapted from natural language processing on the language of proteins, written in the alphabet of amino acids. In this work, we demonstrate how large language models based on pretrained deep neural networks can be effectively finetuned for controllable generation of protein sequences from several distinct protein families. Using bioinformatic and deep learning-based methods, we show that the model is able to generate high-quality protein sequences that exhibit low similarity to existing proteins.
|
|
Knowledge Management in Project Management in International Companies
Peschová, Nikol ; Běhounková, Olga (advisor) ; Gruber, Jan (referee)
The aim of this bachelor's thesis is to explore the application of knowledge management in the context of project management in international companies. First, the concept of knowledge management and related theoretical foundations such as intellectual capital, knowledge workers, and the concept of a learning organization are introduced. Attention is paid to the strategic approach to implementing knowledge management, its benefits and significance in international organizations. Project management, its specifics in international organizations, risks threatening its success, and the importance of knowledge management in project management are described. Furthermore, the thesis is focused on lessons learned from projects, their identification, storage, and sharing, as well as the methods used in each phase of the lessons learned process. The thesis includes a quantitative survey conducted using an online questionnaire. The aim of the survey is to find out whether large international companies apply knowledge management in project management and whether they capture, store, and utilize lessons learned from projects. The respondents of the survey are project managers from the Prague branches of large international companies. Keywords: knowledge management, knowledge economy, knowledge worker,...
|
|
Writing assistant based on large language models
Klement, David ; Helcl, Jindřich (advisor) ; Libovický, Jindřich (referee)
A standard approach to many natural language processing tasks is to take an existing, pre-trained large language model and fine-tune it for the given task. Such an approach leads to having a separate model for each task; furthermore, the fine-tuning must be repeated when upgrading to a new pre-trained model. This thesis explores the possibilities of using a single off-the-shelf model for three different tasks without fine-tuning. We present Preditor, a writing assistant that supports rewriting a sentence after replacing one of its words, suggesting continuations, and suggesting words that fit into a sentence. We design the system in a model-agnostic way, making it possible to upgrade to a new model with little effort. We also provide an extension that integrates the assistant into the text editor. 1
|
|
From Modular to End-to-End Speaker Diarization
Landini, Federico Nicolás ; Ganapathy, Sriram (referee) ; Bredin, Herve (referee) ; Burget, Lukáš (advisor)
Diarizace mluvčího se obvykle popisuje jako úloha, která určuje, "kdo kdy mluvil" v nahrávce. Ještě před několika lety byly všechny konkurenceschopné přístupy modulární, tj. detekce hlasové aktivity, segmentace, extrakce embeddingů, shlukování a detekce a zpracování překrývající se řeči byly řešeny různými subsystémy a aplikovány jeden po druhém. Takto konstruované systémy dosahovaly ve většině scénářů nejlepších výsledků, ale měly velké potíže vypořádat se s překrývající se řečí. V poslední době vzbudil velkou pozornost nástup end-to-end modelů, které jsou schopny řešit všechny aspekty diarizace mluvčího pomocí jediného modelu a mají lepší úspěšnost, pokud jde o překrývající se řeč. Tato práce vznikla v období koexistence těchto dvou trendů. Popisujeme systém založený na bayesovském skrytém Markovově modelu používaném ke shlukování x-vektorů (embeddingů mluvčích získaných pomocí neuronové sítě), známý jako VBx, který prokázal pozoruhodnou úspěšnost na různých souborech dat a evaluačních v různých kampaních. Popíšeme jeho výhody a omezení a vyhodnotíme výsledky na různých relevantních korpusech. Poté přejdeme k metodám end-to-end neuronové diarizace (EEND). Vzhledem k potřebě velkých trénovacích sad pro trénování těchto modelů a nedostatku ručně anotovaných diarizačních dat v dostatečném množství spočívá kompromisní řešení v umělém generování trénovacích dat. Popíšeme přístup ke generování syntetických dat, která se z hlediska změn a překryvů mluvčích podobají skutečným rozhovorům. Ukážeme, jak tato metoda generování simulovaných konverzací umožňuje dosáhnout lepší úspěšnosti než dříve navržená metoda vytváření simulovaných směsí při trénování populárního EEND s enkodér-dekodér atraktory (EEND-EDA). Navrhneme také nový model založený na EEND, který nazýváme DiaPer, a ukážeme, že může dosáhnout lepších výsledků než EEND-EDA, zejména při práci s mnoha mluvčími a při zpracování překrývající se řeči. Nakonec porovnáváme oba systémy, VBx i DiaPer, na široké škále datových sad a prodiskutejeme výhody každého z nich.
|
|
Detection of Similar Binding Sites in Protein Structure Databases
Telčer, Jakub ; Hoksza, David (advisor) ; Galgonek, Jakub (referee)
The evaluation of protein-ligand binding site similarity is crucial in many fields, from drug repurposing trials to evolutionary studies. Current state-of-the- art methods achieve good results on the benchmarking datasets. However, the current approaches operate over pairs of binding sites and are not applicable for searching databases of unprocessed protein structures. In cases when the binding sites are unknown, they have to be firstly located by using binding site prediction algorithms. That significantly increases the upfront costs of creating large databases of similar binding sites. This work covers the current methods for assessing binding site similarity and explores the possibility of fast searching of large databases of related structures by presenting a simple method that allows faster than linear search without the need for identification of the precise locations of the putative binding sites. The proposed approach shows promising preliminary results that merit further investigation, although more insight is still needed.
|
|
Analysis of language of the Formula 1 commentators with main focus on vocabulary
Hais, Petr ; Schneiderová, Soňa (advisor) ; Trunečka, Ondřej (referee)
The thesis covers the topic of sports commentary of Formula 1in Czech and Slovak language. Its goal is to find characteristics of commentary of this sport and compare fundamental differences from sports of the Czech mainstream (football, ice hockey). There is language analysis used for the purpose of the thesis with accent on vocabulary and its usage in the commentary with regard on the difference between vocabulary specific for motorsport and Formula 1 and common vocabulary used by the commentator. The analysis contains eight grand prix from 2021 to 2024. In selecting these races, attractiveness from both perspectives was taken into account, specifically both really attractive and less attractive races were chosen. The definition of specifics and contrasts of the Formula 1 commentary vocabulary is the main goal of the thesis. The text introduces main distinctions between sports journalism and other parts of journalism in combination with an outline of the definition of sports commentary and its problematic in academic world. The results of the thesis are finding out the accuracy and mistakes in usage of words from the Formula 1 vocabulary and a presentation of the intensity of words used in the commentary. In comparison with sports of the Czech mainstream the thesis found out key difference in the...
|