Novel method of identifying DNA methylation fingerprint of acute myeloid leukaemia

Finding new statistical approaches to high throughput data analysis is a very hot topic nowadays. Such a data needs dedicated methods and algorithms of analysis due to huge number of features, but often also due to a small number of samples. Methylation data are also very special, because of dependencies between features and their neighbourhood. There is a need to find a novel, data driven algorithm for these data owing to big variety of distributions data sets. Purpose of this method is detection of regions with different levels of demethylation. From the biological point of view, the most important genome regions are TSS (transcription start site) regions. Hypermethylation of these part of a gene leads to repression and thus stop the gene expression. This phenomenon often happens in cancer disease and impairs a number of molecular processes in the cell. The proposed algorithm is performed for AML patients data in comparison to healthy control. By combination of statistics methods and mathematical modelling together, it enables detection of demethylated regions or DNA and their classification as low, medium or high demethylated.