DATA SETS

Sets of extremely conserved noncoding regions (pvalue <= 1e-40)
identified by Gumby in a whole-genome alignment of human (hg16
assembly), mouse (mm4) and rat (rn3) at 6 different settings of the
Gumby R-ratio parameter (R = 5, 15, 50, 200, 1000, 10000).

The files contain coordinates of conserved elements in the human hg18
genome assembly.

See the first reference below for a complete description of the
method.


REFERENCES

These data sets are described in and form the basis of: Visel A. et
al., Nature Genetics 2008 (Note: the set with R=50 was the primary
data set used in this paper).

The whole-genome alignments were performed using the VISTA genome
alignment pipeline [Frazer KA et al., Nucleic Acids
Res. 2004;32:W273-9] and MLAGAN global alignment program [Brudno M. et
al, Genome Research 2003;13(4):721-31].

Genome sequences were downloaded from the UCSC Genome Bioinformatics
website [http://genome.ucsc.edu/].


FORMAT

The files are in 4-column format:

chr start end log10(1/pvalue)

As in the UCSC BED format, the start-coordinate is zero-based and the
end coordinate is one-based. For example,

chr10:10000-20000

would be written as

chr10	9999	20000

in this format.

The conservation pvalue was calculated by Gumby (Prabhakar S. et al,
Genome Research 2006). The value in the 4th column can be thought of
as a conservation score, i.e., the higher the value, the greater the
degree of evolutionary conservation.


CAVEAT

These lists have some false negatives, in regions where one of the
compared genomes has an incomplete/incorrect assembly, or in regions
that are not syntenically aligned in all three species by the genome
alignment pipeline.

Also, 1e-40 is an extremely strict pvalue threshold, designed to
select only the extreme tail of the distribution of constrained
sequences in the human genome.