Web-based resources for comparative genomics.

The available web-based genome data and related resources provide great opportunities for biomedical scientists to identify functional elements in a particular genome region or to explore the evolutionary pattern of genome dynamics. Comparative genomics is an indispensable tool for achieving these goals. Because of the broad scope of comparative genomics, it is difficult to address all of its aspects in this short survey. A few currently 'hot' topics have therefore been selected and a brief review of the availability of web-based databases and software is given.

Genome databases for comparative genomics

Usually, genome-wide databases (see Table 1) change rapidly, both in their internal implementation and in the datasets recorded. This paper briefly reviews two severs recently made public, which researchers should find valuable for obtaining a wealth of useful information. The genome alignment and annotation database (GALA)[1] provides access to information on genes (known and predicted), gene ontology, expression patterns, genome alignments and conserved transcription factor binding sites predicted by the TRANSFAC weight matrix that can be estimated from the known binding sites to show the sequence signature [2]. For example, given a set of genes expressed in a particular tissue, GALA is able to identify all of the predicted binding sites for one or more transcription factors of interest that are all conserved in mammals. EnsMart is a branch of the Ensembl project,[3] which integrates data from Ensembl and several other resources, using a 'warehouse star-schema' with central biological objects (eg genes or single nucleotide polymorphisms) connected to a set of satellite tables, such as disease, transcript and protein family (PFAM) attributes. Thus, EnsMart provides users with fast and effective access to deep data in and around genes.

Table 1

Websites for tools and databases useful for comparative genomics

Tool or database	Website
NCBI	http://www.ncbi.nlm.nih.gov
ENSEMBL	http://www.ensembl.org
UCSC Genome Browser	http://genome.uscs.edu/
EnsMart	http://www.ensembl.org/Multi/martview
NCBI BLAST	http://www.ncbi.nlm.nih.gov/BLAST/
WU-BLAST	http://blast.wustl.edu/
GALA	http://gala.cse.psu.edu/
PipMaker and MultiPipMaker	http://bio.cse.psu.edu/pipmaker/
zPicture	http://zpicture.dcode.org/
VISTA	http://www-gsd.lbl.gov/vista/
MAVID	http://baboon.math.berkeley.edu/mavid/
MEME	http://meme.sdsc.edu
GLASS and ROSETTA	http://crossspecies.lcs.mit.edu/
SGP2	http://genome.imim.es/software/sgp2/
TWINSCAN	http://genes.cs.wustl.edu/query.html
GeneID	http://www1.imim.es/geneid.html
DOUBLESCAN	http://www.sanger.ac.uk/Software/analysis/doublescan/
TRED	http://rulai.cshl.edu/TRED
RNAdb	http://research.imb.uq.edu.au/rnadb/
NONCODE	http://noncode.bioinfo.org.cn
PAML	http://abacus.gene.ucl.ac.uk/software/paml.html
DIVERGE	http://xgu.zool.iastate.edu
Mgenome	http://xgu.zool.iastate.edu
GRIMM	http://www-cse.ucsd.edu/groups/bioinformatics/GRIMM/
GRAPPA	http://www.cs.unm.edu/~moret/GRAPPA/
TRANSFAC	http://www.gene-regulation.de/
FootPrinter and PhyME	http://bio.cs.washington.edu/software.html
MSARI	http://theory.csail.mit.edu/MSARi/
RNAZ	http://www.tbi.univie.ac.at/~wash/RNAz/

Multi-genome alignment and gene prediction

Genome-wide alignment servers for two closely related species are available on the web. The BLAST,[4,5] implemented at the National Center for Biotechnology Information (NCBI), is the most frequently used suite of tools. Several servers were specially designed to align two or more long genomic sequences at high sensitivity while detecting common rearrangements or duplications -- for example, PipMaker,[6] MultiPipMaker,[7] zPicture,[8] VISTA [9] and MAVID [10]. These servers are suitable for species such as those from different mammalian orders. Several pipelines have been designed for mammalian genome alignment [11-13]. For more distant species, or ancient paralogous genes, different alignment methods should be recommended. One major application is to look for common motifs in the upstream regions of co-expressed genes. Two examples of these approaches are multiple expectation maximisation for motif elicitation (MEME) and Gibbs sampling [14-16].

One application of multi-genome alignment is to improve the efficiency of gene finding. ROSETTA reconstructs co-linear gene structures from global alignments and defines exons as sub-sequences bounded by splice sites [17]. Syntenic Gene Prediction version 1 (SGP18) reconstructs genes from a collection of local alignments between two sequences,[18] while SGP2 assesses the reliability of gene models predicted by GeneID,[19] a conventional gene predictor [20]. Similarly, TWI-NSCAN represents a direct extension of the Genscan algorithm that integrates conservation information between two sequences [21-23]. DOUBLESCAN uses a pair hidden Markov model (Pair HMM) to reconstruct gene structures from a series of local alignments created with BLAST [4,24].

Evolutionary approaches to protein function detection

Phylogenetic analysis by maximum likelihood (PAML) is a software package that includes a wealth of methods for statistically testing the evolutionary pattern of coding sequences, which can be used for one functional detection and prediction of proteins [25]. For instance, PAML is able to estimate ω, the ratio of the non-synonymous rate to the synonymous rate at each amino acid residue along the lineages of a given phylogenetic tree. DIVERGE is a program for studying one functional divergence of a protein family by detecting site-specific changes in the evolutionary rate using a multiple alignment of amino acid sequences for a given phylogenetic tree [26,27]. It first conducts a statistical test for site-specific rate shifts along the tree and predicts candidate amino acid residues responsible for functional divergence based on posterior analysis. These results can then be mapped on the three-dimensional protein structure, if available.

Multiple genome rearrangement by signed reversal

For comparative gene mapping, it is important to reconstruct the ancestral gene orders for given current genomes. Mathematically, it becomes a problem of signed reversals -- that is, how the genomes evolve from a common ancestral genome based on signed reversal of genes or gene sets. Since this problem is now-deterministic polynomial-time hard (NP-hard),[28] most work is focused on heuristic algorithms for reconstructing the gene order of ancestral genomes. Sankoff et al. [29] searched for the optimal ancestral genome for a median problem upon a grid. Bourque and Pevzner [30] designed the model generative reasoning (MGR) algorithm to reconstruct ancestral genomes using a greedy-split strategy. Wu and Gu [31,32] improved the searching accuracy by using a nearest path search algorithm; they developed a neighbour-perturbing algorithm to reconstruct the optimal gene order of ancestral genomes.

Comparative microarray analysis

Because of the limited data available, there are only a few case studies for interspecies microarray analysis. One good example is for the human-chimpanzee expression profile comparisons in the brain and liver [33,34]. Gu [35] developed a statistical framework for studying expression divergence between duplicate genes, which can also be used to infer the ancestral expression profiles when the phylogeny of duplicate genes is known. To facilitate application of these models to expression and genomic data, Gu et al. [36] defined an additive expression distance between duplicate genes, measured by the average of squared expression differences. They analysed yeast gene families using a multi-microarray dataset and found a more than ten-fold increase in the rate of expression evolution immediately following gene duplication.

Identification of functional non-coding elements by comparative genomics

Although the majority of eukaryote genomes are non-coding regions and were previously regarded as 'junk DNA', recent studies have indicated that non-coding regions harbour important functional elements such as cis-regulatory modules [37,38]. Computational detection of these functional non-coding elements has been extremely challenging. It has been recognised that comparative genomics may be a promising approach to solving this problem. 'Phylogenetic footprinting' focuses on the discovery of novel regulatory elements based on the sequence conservation among a set of ortho-logous non-coding regions [39]. Using this method, many successful motif discovery programs have been developed; for example, Gibbs sampler,[40] MEME,[41] Consensus,[42] AlignAce,[43] ANN-Spec,[44] FootPrinter [45] and PhyMe [46]. For non-coding RNA elements, many tools have been developed to identify the evolutionary conservation of secondary structures, such as QRNA,[47] DDBRNA,[48] MSARI,[49] and RNAZ [50]. The development of these tools serves as compelling evidence for biologically relevant non-coding RNAS function. In addition, some databases of functional non-coding elements are also available; for example, TRED,[51] RNAdb [52] and NONCODE [53].

Conclusion

In summary, this paper has briefly reviewed the web-based resources for comparative genomics. Given that substantial resources are available, the challenge in fact turns on how to transfer the explosion in genomic data to biological knowledge. The internet has substantially facilitated the transition process but progress depends on the development of new ideas and analysis pipelines that combine many approaches, including comparative genomics.