Statistical algorithms in the study of mammalian DNA methylation

摘要: DNA methylation is a dynamic chemical modification that abundant on sequences and plays central role in the regulatory mechanisms of cells. This can be inherited across cell divisions generations, providing ``memory mechanism for programs more flexible than coded sequence. In recent years, high-throughput sequencing technologies have enabled genome-wide annotation methylation. Coupled with novel computational machinery, these developments unperceivable insight to characteristics, biological function disease association this phenomenon. The collaborations between experimental researches who take part efforts has been closer ever before due need involve methodologies throughout entire research pipeline, from design through bias correction analysis large datasets. first thesis we present contributions field We introduce statistically sound criteria detection signatures sequence, an algorithm informative non-overlapping subset such regions optimal under biologically motivated assumptions. Our method outputs sequence-generated list are interest respect their states. then Bayesian network infer corrected site-specific states favorable but biased method, describe its incorporation software package. Along calls our package annotates experiment-specific by considering both state inferences genomic These serve as basis comparative studies. last chapter section bring results genome-scale study conducted humans, chimpanzees orangutan, evidence differences propagate generations distinguish closely related species. second concerns error course studying discovered systematic false-positive variant significantly affect variety classifier correct errors show it performs very well sensitivity specificity.