Computational screening of conserved genomic DNA in search of functional noncoding elements

The sequencing of the mouse genome allowed, for the first time, the large-scale estimation of the extent of sequence conservation within our own genome. In particular, it suggested that in mammals there is at Least twice as much conserved genomic DNA as there is protein coding DNA(1). The abundance of conserved noncoding regions holds even for so-called ultraconserved elements at the very tip of the mammalian conservation scale(2), as well as for sequences conserved between long-diverged vertebrates such as human and fish(3,4). Much of the observed conservation appears to be the result of purifying selection, suggesting a wealth of uncharacterized functional elements and families(5), including transcriptional and post-transcriptional regulatory elements, chromatin structure-associated regions, noncoding RNAs, and perhaps altogether novel classes of functional etements(6). Moreover, recent comparative sequencing efforts have revealed similarly rich sets of uncharacterized conserved noncoding sequences in other metazoans(7). We outline here how to obtain sets of conserved regions from a wide range of model organisms. We then describe how to analyze the properties of these regions, filter out undesired ones, such as known and predicted coding regions, and rank the remainder for further computational and functional analysis. The method makes heavy use of the UCSC Genome Browser Database(8) and a suite of related web-accessible tools. The protocol is divided into four major steps: defining the genomic region of interest, based on the user's starting point (gene of interest, a region between two genetic markers, and other regions); selecting a subset of cross-species conserved elements within this region, based on a hidden Markov model that defines and scores genomic intervals for conservation; mapping the different properties of the interval set (such as transcript overlap and species coverage extent); and, finally, ranking the set for further analysis, based on a characteristic profile of the functional class of interest. The same protocol may be used to search for different functional classes of elements in all branches of the tree of life available in the UCSC Genome Browser, including vertebrate, insect, nematode and yeast. It can also easily incorporate custom types of information that the user has access to and allows for easy replacement of parts of the protocol, as our understanding of the relationship between function and sequence conservation, and of the different functional classes, improves. As an example, we present an informatic profile of vertebrate enhancer sequences and discuss a case for which such a method has led to the discovery of several functional enhancers. An accompanying protocol describes a complementary approach to identification of cis-regulatory DNA regions in complex genome assemblies by clustering of sequence motifs corresponding to known transcription factor binding sites(9).