|
|
Methods for Predicting Drug Side Effects in Silico
Cicková, Pavlína ; Lexa,, Matej (referee) ; Berka,, Karel (referee) ; Provazník, Ivo (advisor)
Vývoj a výzkum léčiv je oblastí současné vědy, jejíž nedílnou součástí je i využití výpočetních metod. Z důvodu nákladnosti a časové náročnosti laboratorních přístupů, metody in silico sehrávají svou významnou roli. I přes rychlý vývoj výpočetních technik využívaných při vývoji léků, však není drtivá většina zkoumaných molekul v procesu vývoje úspěšná a do schvalovací fáze nepostoupí. Nejen proto se nejmodernější strategie návrhu potenciálních nových léčiv zaměřují na opětovné zkoumání již schválených léků a berou do úvahy i analýzu podobností. Tato práce popisuje vývoj a aplikaci souboru několika workflow, jež byl vytvořen v rámci analytické platformy KNIME a jež implementuje metody strojového učení za účelem predikce nežádoucích účinků léčiv. Součástí prezentovaných workflow je získání dat, jejich předzpracování, výpočet metrik podobností a provedení explorační analýzy. Následně je využito klasifikačních modelů k predikci specifických nežádoucích účinků léčiv. Tato predikce vychází z principů technik založených na podobnosti. K natrénování modelů rozhodovacích stromů pro predikci potenciální asociace nežádoucích účinků s léčivy byly využity strukturní a jiné podobnosti schválených molekul léčiv. Hlavní přínos práce spočívá především v přenositelnosti použitých metod. Soubor workflow je určen k využití jako vhodný nástroj k řešení výzkumných otázek ohledně podobnosti léčiv a jelikož analytická platforma KNIME poskytuje uživatelsky přívětivé grafické rozhraní, není nutné, aby měli uživatelé pokročilé zkušenosti v oblasti strojového učení nebo programování, aby mohli soubor navržených workflow v rámci této platformy pro své analýzy využít.
|
|
|
Modelling of the interaction of proteins and peptides with metal ions
Gutten, Ondrej ; Konvalinka, Jan (advisor) ; Obšil, Tomáš (referee)
Modelling of interactions of proteins and peptides with metal ions Ondrej Gutten - Diploma thesis Keywords: Metalloproteins, metal ion selectivity, in silico prediction Abstract: An approach for in silico prediction and estimation of selectivity properties of metal-binding peptides is suggested. An in-depth analysis is performed to disclose the justifiability and limitations of this approach. The study is divided into three parts. First part investigates the soundness of two quantum chemical methods (MP2 and DFT) for their use in the set-up quest. The testing includes comparison with CCSD(T), effect of basis selection, performance of the two methods in geometry optimizations and effect of implicit solvent model. Second part foreshadows the approach of searching for a metal selective peptide by thoroughly investigating the ability of simple representative systems, derived from their metalloprotein templates, to retain the property of interest. Final part describes the initial step of extensive combinatorial approach towards examination of vast number of simple systems that represent metal-binding sites, and which are to be used for prediction of metal-selectivity through exploitation of the described approach and, ultimately, to the de novo design of metalloproteins with desired properties.
|
|
|
Experimental verification of in silico predicted protein binder to FOXO4 transcription factor and transcriptome analysis of bladder cancer
Tauš, Petr ; Drbal, Karel (advisor) ; Převorovský, Martin (referee)
This diploma thesis includes an experimental and a bioinformatic part. The two parts are linked together through the subject of transcription factors of 'forkhead box O' (FOXO) family. FOXO transcription factors have a key role in many cellular processes including cell cycle regulation, apoptosis and metabolism. For a long time, they have been considered strictly as the tumor-suppressors yet a growing number of evidence is pointing out to their pro-tumorigenic role. In consequence FOXO transcription factors are studied intensively as potential therapeutic targets in cancer. In the past decade, in silico prediction of protein-protein interactions has become popular in basic research as well as in drug development. Nonetheless, the predicted structures are still far from fitting to the expected behavior of the respective biomolecules. In the experimental part of this thesis, I verified the interaction of four in silico predicted protein binders based on naturally occurring PDZ domain with FOXO4 using microscale thermophoresis. Non-invasive bladder tumors represent a heterogeneous disease where reliable prediction of tumor aggressiveness is still lacking despite an intensive research. In the bioinformatic part of this thesis, I described the cellular composition of the tumor microenvironment and demonstrated...
|
|
|
Fasciolid flukes: from genes to diagnostic tools
Ježková, Monika ; Leontovyč, Roman (advisor) ; Sojka, Daniel (referee)
Liver flukes of the family Fasciolidae are parasites of mammals including human. Fascioloides magna and Fasciola hepatica are considered as a veterinary and medically important species occurring also in the Czech Republic. Fascioloides magna and F. hepatica infect wide spectrum of wild and domestic ruminants and in case of F. hepatica human can be also infected. Both flukes are responsible for damage of liver tissue and/or bile-ducts of their definitive hosts causing weight lose, anemia, reduced productivity and in specific cases the death of the host. Effective diagnosis plays the key role in control of F. hepatica and F. magna infections. Current diagnostics is predominantly based on serodiagnostic methods using specific antigens e.g. from excretory-secretory products (ESPs). Due to heterogenity of ESPs, such diagnostic markers can lack the specificity and also the reproducibility of the method is poor. Particular proteins of ESPs are often used in diagnostics of fasciolid flukes. Such approach requires biological material and laboratory procedures associated with identification, purification and antigenicity testing of selected proteins. Recent development of parallel sequencing technologies results in huge amount of genomic, transcriptomic and proteomic data, which are publicly available. Such...
|
|
|
Experimental verification of in silico predicted protein binder to FOXO4 transcription factor and transcriptome analysis of bladder cancer
Tauš, Petr ; Drbal, Karel (advisor) ; Převorovský, Martin (referee)
This diploma thesis includes an experimental and a bioinformatic part. The two parts are linked together through the subject of transcription factors of 'forkhead box O' (FOXO) family. FOXO transcription factors have a key role in many cellular processes including cell cycle regulation, apoptosis and metabolism. For a long time, they have been considered strictly as the tumor-suppressors yet a growing number of evidence is pointing out to their pro-tumorigenic role. In consequence FOXO transcription factors are studied intensively as potential therapeutic targets in cancer. In the past decade, in silico prediction of protein-protein interactions has become popular in basic research as well as in drug development. Nonetheless, the predicted structures are still far from fitting to the expected behavior of the respective biomolecules. In the experimental part of this thesis, I verified the interaction of four in silico predicted protein binders based on naturally occurring PDZ domain with FOXO4 using microscale thermophoresis. Non-invasive bladder tumors represent a heterogeneous disease where reliable prediction of tumor aggressiveness is still lacking despite an intensive research. In the bioinformatic part of this thesis, I described the cellular composition of the tumor microenvironment and demonstrated...
|
|
|
Modelling of the interaction of proteins and peptides with metal ions
Gutten, Ondrej ; Konvalinka, Jan (advisor) ; Obšil, Tomáš (referee)
Modelling of interactions of proteins and peptides with metal ions Ondrej Gutten - Diploma thesis Keywords: Metalloproteins, metal ion selectivity, in silico prediction Abstract: An approach for in silico prediction and estimation of selectivity properties of metal-binding peptides is suggested. An in-depth analysis is performed to disclose the justifiability and limitations of this approach. The study is divided into three parts. First part investigates the soundness of two quantum chemical methods (MP2 and DFT) for their use in the set-up quest. The testing includes comparison with CCSD(T), effect of basis selection, performance of the two methods in geometry optimizations and effect of implicit solvent model. Second part foreshadows the approach of searching for a metal selective peptide by thoroughly investigating the ability of simple representative systems, derived from their metalloprotein templates, to retain the property of interest. Final part describes the initial step of extensive combinatorial approach towards examination of vast number of simple systems that represent metal-binding sites, and which are to be used for prediction of metal-selectivity through exploitation of the described approach and, ultimately, to the de novo design of metalloproteins with desired properties.
|
guest ::
