|
|
Methods for Comparative Analysis of Metagenomic Data
Sedlář, Karel ; Vinař,, Tomáš (oponent) ; Lexa, Matej (oponent) ; Provazník, Ivo (vedoucí práce)
Modern research in environmental microbiology utilizes genomic data, especially sequencing of DNA, to describe microbial communities. The field studying all genetic material present in an environmental sample is referred to as metagenomics. This doctoral thesis deals with metagenomics from the perspective of bioinformatics that is unreplaceable during the data processing. In the theoretical part of this thesis, two different approaches of metagenomics are described including their main principles and weaknesses. The first approach, based on targeted sequencing, is a well-established field with a wide range of bioinformatics techniques. Yet, methods for comparison of samples from several environments can be highly improved. The approach introduced in this thesis uses unique transformation of data into a bipartite graph, where one partition is formed by taxa, while the other by samples or environments. Such a graph fully reflects qualitative as well as quantitative aspect of analyzed microbial networks. It allows a massive data reduction to provide human comprehensible visualization without affecting the automatic community detection that can found clusters of similar samples and their typical microbes. The second approach utilizes whole metagenome shotgun sequencing. This strategy is newer and the corresponding bioinformatics techniques are less developed. The main challenge lies in fast clustering of sequences, in metagenomics referred to as binning. The method introduced in this thesis utilizes a genomic signal processing approach. By thorough analysis of redundancy of genetic information stored in genomic signals, a unique technique was proposed. The technique utilizes transformation of character sequences into several variants of phase signals. Moreover, it is able to directly process nanopore sequencing data in the form of a native current signal.
|
|
|
Methods for fast sequence comparison and identification in metagenomic data
Kupková, Kristýna ; Škutková, Helena (oponent) ; Sedlář, Karel (vedoucí práce)
The objective of this thesis is to create a method for identification of organisms in metagenomic data. Until this point methods based on sequence alignment with reference database have been sufficient for this purpose. However, the volume of data grows rapidly with evolvement of sequencing techniques and the alignment-based methods became inconvenient due to computationally demanding alignment. A new technique is introduced in this master’s thesis, which allows alignment-free metagenomic data classification. The method is based on transformation of sequences to genomic signals in form of phase representation, from which feature vectors are extracted. These features are three Hjorth descriptors, which are then subjected expectation maximization for Gaussian mixture model method allowing reliable binning of metagenomic data.
|
|
|
Binning numerical variables in credit risk models
Mattanelli, Matyáš ; Baruník, Jozef (vedoucí práce) ; Teplý, Petr (oponent)
Tato práce zkoumá vliv diskretizace numerických proměnných na výkonnost modelů kreditního rizika. Rozdíly ve výkonnosti jsou vyhodnoceny s využitím pěti veřejně dostupných datových souborů, šesti indikátorů výkonnosti a stati- stického testu. Výsledky naznačují, že diskretizace má pozitivní a významný vliv na výkonnost logistické regrese, neuronové sítě a naivního Bayes klasi- fikátoru. Nejvíce ovlivněným aspektem výkonnosti modelu se zdá být jeho schopnost rozlišovat mezi dobrými a špatnými klienty. Výsledky jsou zvláště patrné pro středně velké datové soubory. Závěry jsou odolné vůči chybějícím hodnotám, eliminaci extrémních pozorování a vyloučení kategorických proměn- ných. Pro rozhodovací strom a náhodný les nebyl nalezen žádný významný pozitivní účinek diskretizace na výkonnost.
|
|
|
Analysis of single-cell genomic data of Saccinobaculus sp.
Gajdošová, Petra ; Treitli, Sebastian Cristian (vedoucí práce) ; Kolísko, Martin (oponent)
Pokrok v metodách jednobuněčné genomiky a metagenomiky nám umožňuje prozk- oumat nekultivovatelné organizmy mnohem podrobněji. V této práci se zaměřujeme na sestavení genomu a genetický kód druhu Saccinobaculus ambloaxostylus ze skupiny oxymonád, který obývá zadní střevo dřevožravých švábů rodu Cryptocercus. Pomocí celogenomové amplifikace jsme získali sekvenční data tří vybraných buněk S. ambloax- ostylus. Získaná data byla použita k sestavení genomu za použití programu SPAdes. Sestavený genom byl následně rozdělen do přihrádek "košů" a dekontaminován, abychom odstranili potenciální bakteriální nebo eukaryotickou kontaminaci. Po dekontaminaci a opětovném sestavení genomu jsme získali genom s celkovou délkou ∼417Mbp, který byl distribuován mezi ∼185 tisíc fragmentů. I přes to, že jsme získali jenom částečný genom s předpokládanou úplností 13,71%, jsme se pokusili určit genetický kód, který používá S. ambloaxostylus. K tomu jsme vymodelovali 33 genů z různých metabolických drah. Porovnáním genových modelů a konzervovaných úseků u homologických proteinů příbuzných oxymonád jsme odhadli, jaké aminokyseliny kódují kanonické stop kodony. Naše výsledky naznačují, že tento organizmus používá genetický kód č. 6 známý napřík- lad u nálevníků, ve kterém jsou stop kodony UAA a UAG přeloženy jako glutamin....
|
|
|
Methods for Comparative Analysis of Metagenomic Data
Sedlář, Karel ; Vinař,, Tomáš (oponent) ; Lexa, Matej (oponent) ; Provazník, Ivo (vedoucí práce)
Modern research in environmental microbiology utilizes genomic data, especially sequencing of DNA, to describe microbial communities. The field studying all genetic material present in an environmental sample is referred to as metagenomics. This doctoral thesis deals with metagenomics from the perspective of bioinformatics that is unreplaceable during the data processing. In the theoretical part of this thesis, two different approaches of metagenomics are described including their main principles and weaknesses. The first approach, based on targeted sequencing, is a well-established field with a wide range of bioinformatics techniques. Yet, methods for comparison of samples from several environments can be highly improved. The approach introduced in this thesis uses unique transformation of data into a bipartite graph, where one partition is formed by taxa, while the other by samples or environments. Such a graph fully reflects qualitative as well as quantitative aspect of analyzed microbial networks. It allows a massive data reduction to provide human comprehensible visualization without affecting the automatic community detection that can found clusters of similar samples and their typical microbes. The second approach utilizes whole metagenome shotgun sequencing. This strategy is newer and the corresponding bioinformatics techniques are less developed. The main challenge lies in fast clustering of sequences, in metagenomics referred to as binning. The method introduced in this thesis utilizes a genomic signal processing approach. By thorough analysis of redundancy of genetic information stored in genomic signals, a unique technique was proposed. The technique utilizes transformation of character sequences into several variants of phase signals. Moreover, it is able to directly process nanopore sequencing data in the form of a native current signal.
|
|
|
Methods for fast sequence comparison and identification in metagenomic data
Kupková, Kristýna ; Škutková, Helena (oponent) ; Sedlář, Karel (vedoucí práce)
The objective of this thesis is to create a method for identification of organisms in metagenomic data. Until this point methods based on sequence alignment with reference database have been sufficient for this purpose. However, the volume of data grows rapidly with evolvement of sequencing techniques and the alignment-based methods became inconvenient due to computationally demanding alignment. A new technique is introduced in this master’s thesis, which allows alignment-free metagenomic data classification. The method is based on transformation of sequences to genomic signals in form of phase representation, from which feature vectors are extracted. These features are three Hjorth descriptors, which are then subjected expectation maximization for Gaussian mixture model method allowing reliable binning of metagenomic data.
|
host ::
RSS.