PuSH - Publikationsserver des Helmholtz Zentrums München

Development of biomarkers by integration of data and prior knowledge

München, Ludwig-Maximilians-Universität & Technische Universität, Bioinformatik, Bachelor-Thesis, 2014, 46 S.
Verlagsversion Volltext
Biological systems, as complex as they may be, exhibit certain behavioural patterns as a response to certain conditions. Although each system’s pattern on the same condition differs to a certain extent from the others, it still remains a pattern and one can find analogies, when comparing them with each other. Key players, holding these patterns together, are called biomarkers. Biomarkers are biological measures of a biological system’s state and can be used as indicators or predictors of conditions, be it internal or external. Biomarkers allow us to compare biological systems or processes, in particular, to examine a wealthy processe against a pathogenic processes or pharmacologic response to therapeutic treatment. Based on biomarkers, hypotheses about present or future biological conditions can be made, providing us with crucial biological knowledge, which in turn may serve as the basis of other research, particularly in disease diagnosis and treatment. The discovery of biomarkers is still a challanging task and great efforts were undertaken to develop techniques and methods to approach this problem. To counter this problem, one of these methods (stSVM [10]) integrates biological network information as well as experimental data into one classifier. By smoothing genewise t-statistics over the graph structure of a PPI-network and subsequent classification, it is capable to provide us with accurate and in particular biologically interpretable results with high signature stability. Our approach makes use of this existing method and extends it by two components, an automated network retrievel system and an automated validation system through PCR data. Our datasets mainly derive from former work done by Quaranta et al. [22], where the detection of biomarkers in two common inflammatory skin diseases, psoriasis and eczema, were targeted. Psoriasis and Eczema are two common widespread inflammatory skin diseases, whose phenotypic outcomes are quite similar, thus hampering to clearly differentiate between these two. The high inter-individual variability, partially based upon gender, age and short-term environmental exposure, makes it even harder to get a comprehensive understanding of disease pathogenesis. In addition, psoriasis ad eczema respond quite differently, even antipodal, to particular therapy methods, hence it is of high interest to develop specific therapies or diagnostic tools, in order to combat these diseases. This former research revealed 174 significantly up- or down-regulated genes either in psoriasis, eczema or both and is our primary input, available as microarray data. The second input forms the PCR data, where a few biologically significant genes, manually selected from the set of significantly up- or down-regulated genes, were remeasured for validation. The two main targets of this study are to build an interface to the STRING database in order to fetch protein interaction data to compile biological networks and secondly to asses genes differential expression via gene-wise t-statistics based on microarray data and subsequent correction through gene-wise t-statistics obtained through PCR data. Fortunately, we were able to improve the aforementioned method (stSVM) and hope to alleviate further efforts of biomarker detection, on the basis of our work.
Weitere Metriken?
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Sonstiges: Hochschulschrift
Typ der Hochschulschrift Bachelorarbeit
Quellenangaben Band: , Heft: , Seiten: 46 S. Artikelnummer: , Supplement: ,
Hochschule Ludwig-Maximilians-Universität & Technische Universität
Hochschulort München
Fakultät Bioinformatik