We developed the Genomic Regions Enrichment of Annotations Tool (GREAT) to analyze the functional significance of cis-regulatory regions identified by localized measurements of DNA binding events across an entire genome. Whereas previous methods took into account only binding proximal to genes, GREAT is able to properly incorporate distal binding sites and control for false positives using a binomial test over the input genomic regions. GREAT incorporates annotations from 20 ontologies and is available as a web application. Applying GREAT to data sets from chromatin immunoprecipitation coupled with massively parallel sequencing (ChIP-seq) of multiple transcription-associated factors, including SRF, NRSF, GABP, Stat3 and p300 in different developmental contexts, we recover many functions of these factors that are missed by existing gene-based tools, and we generate testable hypotheses. The utility of GREAT is not limited to ChIP-seq, as it could also be applied to open chromatin, localized epigenomic markers and similar functional data sets, as well as comparative genomics sets.

We have conducted a comprehensive search for conserved elements in vertebrate genomes, using genome-wide multiple alignments of five vertebrate species (human, mouse, rat, chicken, and Fugu rubripes). Parallel searches have been performed with multiple alignments of four insect species (three species of Drosophila and Anopheles gambiae), two species of Caenorhabditis, and seven species of Saccharomyces. Conserved elements were identified with a computer program called phastCons, which is based on a two-state phylogenetic hidden Markov model (phylo-HMM). PhastCons works by fitting a phylo-HMM to the data by maximum likelihood, subject to constraints designed to calibrate the model across species groups, and then predicting conserved elements based on this model. The predicted elements cover roughly 3%-8% of the human genome (depending on the details of the calibration procedure) and substantially higher fractions of the more compact Drosophila melanogaster (37%-53%), Caenorhabditis elegans (18%-37%), and Saccharaomyces cerevisiae (47%-68%) genomes. From yeasts to vertebrates, in order of increasing genome size and general biological complexity, increasing fractions of conserved bases are found to lie outside of the exons of known protein-coding genes. In all groups, the most highly conserved elements (HCEs), by log-odds score, are hundreds or thousands of bases long. These elements share certain properties with ultraconserved elements, but they tend to be longer and less perfectly conserved, and they overlap genes of somewhat different functional categories. In vertebrates, HCEs are associated with the 3Ј UTRs of regulatory genes, stable gene deserts, and megabase-sized regions rich in moderately conserved noncoding sequences. Noncoding HCEs also show strong statistical evidence of an enrichment for RNA secondary structure.

There are 481 segments longer than 200 bp that are absolutely conserved (100% identity with no insertions or deletions) between orthologous regions of the human, rat and mouse genomes. Nearly all of these segments are also conserved in the chicken and dog genomes, with an average of 95% and 99% identity, respectively. Many are also significantly conserved in fish. These ultraconserved elements of the human genome are most often located either overlapping exons in genes involved in RNA processing or in introns or nearby genes involved in regulation of transcription and development. Along with more than 5,000 sequences of over 100bp that are absolutely conserved among the three sequenced mammals, these represent a class of genetic elements whose functions and evolutionary origins are yet to be determined, but which are more highly conserved between these species than proteins, and appear to be essential for the ontogeny of mammals and other vertebrates.Although only about 1.2% of the human genome appears to code for protein (1-3), it has been estimated that as much as 5% is more conserved than expected from neutral evolution since the split with rodents, and hence may be under negative or "purifying" selection (4-6). Several studies have found specific non-coding segments in the human genome that appear to be under selection, using a threshold for conservation of 70% or 80% identity with mouse over more than 100bp (7-13). A study of these elements on human chromosome 21 found that those that were very highly conserved in multiple species contained significant numbers of non-coding elements (13). Similar results were found comparing the human, mouse and rat (14, 15) in a study of the 1.8 Mb CFTR region (16,17), and in a functional study of the SIM2 locus in a number of mammalian species (18). We determined the longest segments of the human genome that are maximally conserved with orthologous segments in rodents: those showing 100% identity and with no insertions or deletions in their alignment with mouse and rat. Exclusive of ribosomal RNA regions, there are 481 such segments longer than 200bp that we call ultraconserved elements (table S1). They are widely distributed in the genome (on all chromosomes except chromosomes 21 and Y), and are often found in clusters (Fig. 1). The probability is less than 10 -22 of finding even one such element in 2.9 billion bases under a simple model of neutral evolution with independent substitutions at each site, using the slowest neutral substitution rate that is observed for any 1 Mb region of the genome (supporting text, section S1). Nearly all of these elements also exhibited extremely high levels of conservation with orthologous regions in the chicken genome (467/481 = 97% of the elements aligning at an average of 95.7% identity, 29 at 100% identity), and about two-thirds of them with the fugu genome as well (324/481 = 67.3% of the elements aligning at an average of 76.8% identity), despite the fact that only about 4% of the human genome can be reliably aligned to the chicken ...