Predicting Methylation Status Of CpG Islands From Genomic Sequences

Epigenetics attends to study the heritable changes in gene expression without altering DNA sequences. DNA methylation is such a most important epigetic mark, which usually exists in the dinucleotide CpG. Over 60% of human genes contain CpG islands in their promoters. The methylation status of these CpG islands plays crucial role in gene transcription. Mis-epigenetic silencing associated with aberrant CpG island methylation is one mechanism leading to the loss of tumor suppressor functions in cancer cells. Thus, to clarify the methylation status is helpful to understand the mechanism of gene regulation. Meanwhile, it is of great use to disease diagnosis and therapy.Large-scale experimental detection of methylation status of CpG islands is still both labor-intensive and time-consuming. Existing research has confirmed that there are some signals embedded in DNA sequences directing the methylation of CpG islands. Therefore, using computational methods to predict methylation status of CpG islands enjoys great concern at present. In this thesis, based on a recent genome-scale dataset of DNA methylation in human brain tissues, we developed a classifier for predicting methylation status of CpG islands using a Support Vector Machine (SVM). Nucleotide sequence contents, transcription factor binding sties and Alu repeats are used as features for the classification. The method achieves accuracy of ~85% on the brain data. We have constructed an on-line server named MethCGI to predict methylation status of CpG islands. Compared with existing methods, we found MethCGI performed outstandingly.Furthermore, we investigated the mechanism buried in DNA sequences to protect methylation-resistant CpG islands from DNA methylation. From the comparison between methylation-resistant and methylation-prone CpG islands, we found some special motifs as cis-elements of zinc-finger proteins, some of which have been validated to function as cis-elements to hold back the spreading of DNA methylation. In the end, we reviewed the relevant research of other epigenetic marks such as nucleosome positioning, histone modification and imprinting genes, including the methods of experimental detection and computational research.