CpG Island Finder — Gardiner-Garden & Takai-Jones Criteria
Find CpG islands by either published definition, with the observed/expected CpG ratio reported separately from GC content.
🔒 Nothing you paste is logged or stored on our servers
Paste a promoter region, a gene or a whole locus and get its CpG islands: the unmethylated, CpG-dense stretches that sit over most vertebrate promoters, and the first thing to establish before designing bisulfite primers or reading a methylation array. Two published definitions are offered because they answer different questions. Gardiner-Garden & Frommer's 1987 thresholds (GC above 50%, observed/expected CpG above 0.6, at least 200 bp) are what most papers mean by “a CpG island”, and are permissive enough to also call many Alu repeats. Takai & Jones tightened them in 2002 (GC at least 55%, obs/exp at least 0.65, at least 500 bp) specifically to exclude Alu and keep the islands that mark promoters. Every island returned satisfies the criteria it is reported under — windows are merged only where the merged region still passes — and the CpG dinucleotide count is reported separately from GC content, since conflating the two is the usual way this number goes wrong.
Paste a sequence to scan it for CpG islands.
How to use the CpG Island Finder tool
- 1Paste your DNA, raw or as FASTA.
- 2Choose a definition: the classic 1987 criteria, or the stricter 2002 ones that exclude most Alu repeats.
- 3Read the island track and table — each island's span, GC%, observed/expected CpG ratio and CpG site count, against the whole sequence's own averages.
Frequently asked questions
How is the observed/expected CpG ratio calculated?
Observed CpG dinucleotides divided by (number of C × number of G) ÷ length — Gardiner-Garden & Frommer's 1987 formula exactly. The key point is what it counts: the dinucleotide CpG read along one strand, never a C followed by anything else, and never the combined C+G of “GC content”. Those are two different numbers and this tool reports both, because confusing them shifts the answer by roughly a factor of two.
Which definition should I use?
Use Takai & Jones 2002 if you want the islands that mark promoters: it was written specifically to exclude Alu repeats, which the 1987 criteria also call. Use Gardiner-Garden & Frommer 1987 if you need to match what most published work means by a CpG island, or to compare against an older analysis. Neither is wrong; they draw the line in different places, and the thresholds used are shown with every result.
Why is an island shorter than my window not reported?
A region has to meet all three thresholds — GC, observed/expected ratio, and minimum length — to count. Regions that pass the first two but fall short on length are counted in the notes rather than silently dropped, so you can see whether lowering the length threshold would find what you are looking for.
More
Related tools
Read a GFF3 annotation file as a feature table, and turn it plus its FASTA into a GenBank record with an annotated map.
Generate composition, ORF, restriction-site and primer summaries from one sequence.
Get the reverse complement, complement and reverse of a DNA/RNA sequence at once.