newcpgreport identifies CpG islands in one or more nucleotide sequences. The ratio of observered to expected number of GC dinucleotides patterns is calculated over a window (sequence region) which is moved along the sequence. The calculated ratios are used to identify regions which match the program's definition of a "CpG island" (a CG dinucleotide rich area). A report file is written giving the input sequence name, CpG island parameters and data on any CpG islands that are found.
The ratio of observered to expected number of GC dinucleotides patterns is calculated over a window of user-specified size (-window parameter). The window is slid along the sequence and the ratio recalculated until the end of the sequence is reached.
By default, cpgplot defines a CpG island as a region where, over an average of 10 windows and not less than 200 bases, the calculated (%G + %C) content is over 50% and the calculated Observed/Expected ratio is over 0.6. These conditions can be modified by setting the values of the appropriate parameters.
The Observed number of CpG patterns in a window is simply the number of times a C is found followed immediately by a G.
The Expected number of CpG patterns is calculated for each window as the number of CpG dinucleotides you would expect to see in that window based on the frequency of C's and G's in that window. Thus, the Expected frequency of CpG's in a window is calculated as the number of Cs in the window multiplied by the number of Gs in the window, divided by the window length. Expected = (number of C's * number of G's) / window length
|
|
"CpG" refers to a C nucleotide immediately followed by a G. The 'p' in 'CpG' refers to the phosphate group linking the two bases. Regions of genomic sequences rich in the CpG pattern or "CpG islands" are resistant to methylation and tend to be associated with genes which are frequently switched on. It's been estimated that about half of all mammalian genes, and, possibly all mammalian house-keeping genes, have a CpG-rich region around their 5' end. Non-mammalian vertebrates have some CpG islands that are associated with genes, but the association gets equivocal in the farther taxonomic groups. The detection of CpG island upstream of predicted exons or genes is evidence in support of a highly expressed gene.
newcpgreport is used in the production of the CpG Island database 'CPGISLE'. It produces a report for potential CpG islands in CPGISLE database entry format. See the FTP site: ftp://ftp.ebi.ac.uk/pub/databases/cpgisle/ for the finished database.
As there is no official definition of what is a CpG island is or how to identify where they begin and end, we work with two definitions and thus two methods. These are:
1. cpgplot and newcpgreport use a sliding window within which the Observed/Expected ratio of CpG is calculated. For a sequence region to reported as a CpG island, it must satisfy the following contraints:
Observed/Expected ratio > 0.6 % C + % G > 50% Sequence Length > 200
2. newcpgseek and cpgreport use a running sum calculated from all positions in a sequence rather than a window to produce a score. If there is not a CG dinucleotide at a position, the score is decremented, if there is one, the score is incremented by a constant (user-defined) value. If the score for a region in the sequence is higher than a threshold (17 at the moment) then a putative island is declared. Sequence regions scoring above the threshold are searched for recursively.
This method overpredicts islands but finds the smaller ones around primary exons. newcpgseek uses the same method as cpgreport but the output is different and more readable. For most purposes you should probably use newcpgreport rather than cpgreport. It is used to produce the human cpgisland database you can find on the EBI's ftp server as well as on the EBI's SRS server.
newcpgseek and cpgreport both now display the actual CpG count, the (%C + %G) and the Observed/Expected ratio in the region where the score is above the threshold.
The geecee program measures CG content in the entire input sequence and is not to be used to detect CpG islands. It can be useful for detecting sequences that MIGHT contain an island.
As there is no official definition of what is a cpg island is, and worst where they begin and end, we have to live with 2 definitions and thus two methods. These are:
1. newcpgseek and cpgreport - both declare a putative island if the score is higher than a threshold (17 at the moment). They now also display the actual CpG count, the % CG and the observed/expected ration in the region where the score is above the threshold. This scoring method based on sum/frequencies overpredicts islands but finds the smaller ones around primary exons. newcpgseek uses the same method as cpgreport but the output is different and more readable.
2. newcpgreport and cpgplot use the method which mentioned in the Description section above. The important thing to note in this method is that an island, in order to be reported, is defined as a region that satisfies the following contraints:
Obs/Exp ratio > 0.6 % C + % G > 50% Length > 200.
For all practical purposes you should probably use newcpgreport. It is actually used to produce the human cpgisland database you can find on the EBI's ftp server as well as on the EBI's SRS server.
geecee measures CG content in the entire input sequence and is not to be used to detect CpG islands. It can be useful for detecting sequences that MIGHT contain an island.