An evaluation of new criteria for CpG islands in the human genome as gene markers

摘要

Motivation: Recently, more stringent criteria for CpG islands have been introduced to exclude Alu repeats, thereby enabling a higher proportion of CpG islands associating with genes to be identified. Using these new criteria, several types of associations between CpG islands and genes were investigated to further establish the importance of CpG islands as gene markers. Results: The CpG islands were searched by CpGIE, a java software program developed for CpG island identification. CpGIE was advanced in identification accuracy compared with other tools. According to our results, about 70% of the identified CpG islands were associating with the human genes and over half of them are in the promoters. Furthermore, the investigation of genes in the confirmed gene model showed that 56% of them had a CpG island overlapping the transcription start sites. In comparison, the new criteria were found capable of filtering a large fraction of Alu repeats that was identified as CpG islands by the generally accepted criteria within the genes, but very few CpG islands associating with the promoters were affected. The genes in the predicted gene model were not obviously associated with CpG islands, suggesting that CpG islands can be used to evaluate the accuracy of gene annotation.