![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
|
RESOURCES GALA, a Database for Genomic Sequence Alignments and AnnotationsDepartments of 1Computer Science and Engineering,2 Biochemistry and Molecular Biology, 3Biology, and4 Huck Institute for Life Sciences, The Pennsylvania State University, University Park, Pennsylvania 16802, USA
We have developed a relational database to contain whole genome sequence alignments between human and mouse with extensive annotations of the human sequence. Complex queries are supported on recorded features, both directly and on proximity among them. Searches can reveal a wide variety of relationships, such as finding all genes expressed in a designated tissue that have a highly conserved noncoding sequence 5' to the start site. Other examples are finding single nucleotide polymorphisms that occur in conserved noncoding regions upstream of genes and identifying CpG islands that overlap the 5' ends of divergently transcribed genes. The database is available online at http://globin.cse.psu.edu/ and http://bio.cse.psu.edu/.
The determination and annotation of complete genomic DNA sequences provide the opportunity for unprecedented advances in our understanding of evolution, genetics, and physiology, but the amount and diversity of data pose daunting challenges as well. Three excellent browsers provide access to the sequence and annotations of the human genome, viz., the human genome browser (HGB) at UCSC (Kent et al. 2002  ) (http://genome.ucsc.edu/), Map Viewer at the National
Center for Biotechnology Information (NCBI)
(http://www.ncbi.nlm.nih.gov/), and Ensembl (Hubbard et al. 2002) at
the Sanger Centre (http://www.sanger.ac.uk/) and EBI
(http://www.ebi.ac.uk/). These sites show known and predicted genes,
repetitive elements, genetic markers, and many other types of
information as separate tracks in a display using the coordinate system
of the human genomic DNA sequence. Thus, users can visually integrate
information from all these sources. However, browsers show a single
gene or region at a time. They are not designed to support complex
queries across multiple types of information simultaneously or multiple
genes. For instance, one cannot easily find all genes encoding
zinc-finger proteins that are expressed in lung cells, or single
nucleotide polymorphisms (SNPs) that occur within a specified distance
of such genes. In contrast, recording appropriate annotations in a
relational database allows users to query multiple features, thereby
supporting searches for nonobvious relationships. The need for greater
data integration is leading to new data-mining capacities, such as
EnsMart at Ensembl.
Alignments of genomic DNA sequences between species at an appropriate
phylogenetic distance are useful for finding candidate functional
regions (Hardison and Miller 1993
We have developed a database of genomic DNA sequence alignments and
annotations, called GALA, to search across tracks of data supplied by
current browsers, alignment resources, and databases. Output from GALA
queries can be viewed as tracks on the Human Genome Browser, as a table
of data with hyperlinks, or as text. When users of the GALA database
wish to view alignments, they can choose to display their query output
using local alignments with Java (Laj), which is a versatile,
interactive alignment viewer implemented using Java (Wilson et al.
2001
The GALA database records information about human genomic DNA (such as genes, repeats, SNPs, CpG islands, recombination frequencies), the structure and function of the products encoded by the genes (including biological process, molecular function, conserved domains, expression patterns), pathological consequences of mutations, and similarity between human and mouse genomic DNA sequences (Table 1). GALA makes extensive use of publicly available datasets, and it incorporates the results of alignments of the human and mouse genomes (Waterston et al. 2002 ; Schwartz et al.
2003). GALA complements and extends the current resources by allowing
simple and complex queries across multiple types of information,
effectively integrating evolutionary insights (alignments) with
functional and structural annotations. Often, users wish to know about
features within a certain distance of other features, rather than
specifying precise coordinates. To accomplish this, GALA allows queries
for features that are in proximity to or cluster with other features.
Query Page The fields available for querying appear on a single page organized as scrollable list boxes, drop-down menus, fill-in text boxes, and/or check boxes for each type of information (examples are in Fig. 1A). Having users choose entries from list boxes and drop-down menus for most fields enforces a controlled vocabulary. Many of the fields can be searched for all values by clicking a check box, which is a shortcut to highlighting every item in the field. Fill-in boxes are used to enter coordinates, and they are employed in cases where the list of choices is too long for a scrollable list to be practical (e.g., lists with over 1000 choices). URL hyperlinks adjacent to data fields on the query page point the user to the original source of the data.
Output Formats GALA provides a variety of choices for the output format of the results. The "region summary" and "gene summary" produce index pages (Fig. 1B) from which one can further explore the selected regions, examining overlap with any or all of the categories available: genes, gap-free alignments, local alignments, SNPs, repeats, CpG islands, or transcription-factor binding sites. Clicking on the "View all data for ..." button (Fig. 1B) will bring up a page with the information selected (via the check-boxes) for the range returned as the result of a query (Fig. 1C). Plain-text output is provided for off-line examination and importing into other programs.
Three graphical displays are available. The first is a custom user
track on the UCSC Human Genome Browser (Kent et al. 2002
Simple Queries Simple queries are made using one or more fields from the query page. If more than one field is chosen, the output will be an intersection of information from the data tables. An example of a simple query specifies "Find all genes encoding zinc-finger proteins that are expressed in lung tumors or cancers." The user chooses "Zinc finger", "Zinc finger domain" and related terms from the list box for "Conserved domain", and then "lung (tumor)", "lung (carcinoma)", and related terms from "Tissue expressed in", as illustrated in Figure 1A. This finds 89 genes, which are listed on the results page; a partial list of genomic regions containing these is shown in Figure 1B. (The number of records returned is subject to change as the database tables are updated with more information.) One can select a view by genes instead of regions on the query page. The user selects the classes of fields to view on the results page (Fig. 1B). Clicking on the button to "View all data for chr 1 152,802,808 to 152,805,338" returns the requested information, showing that the region contains ZFP67 (c-Krox), which is a Zn finger gene expressed in lung tumors (as expected; Fig. 1C). The results include much additional information, for example, about gene structure, expression in other tissues, proposed functions, and other conserved domains in the encoded protein, including hyperlinks to more information (Fig. 1C).
Complex Queries and the History Page The first example of a complex query is to search for SNPs in conserved noncoding sequences located proximal to the 5' ends of genes. Many regulatory regions are conserved above the background level of sequence identity in a region, and one may want to search for those that are close to the 5' ends of genes. Furthermore, it is possible that some SNPs in conserved noncoding sequences could contribute to changes in gene expression. Thus, this query will return candidate regulatory elements that may be involved in alterations in gene expression.
The complex query begins with three simple queries. First, one asks for
all the known genes in the database (currently 33,061 entries). Then,
one finds all gap-free alignments between human and mouse of 100 bp in
length and 70% or more in identity, which are the criteria used by
Loots et al. (2000)
These three simple queries are then filtered and combined on the
history page to produce a dataset that answers the question of
interest. First, we restrict alignments to those in proximity to genes;
in particular, results from query 2 (highly conserved gap-free
alignments) that are within 2000 bp upstream of results from query 1
(all genes) are retained; this is 22,894 regions. Next, genes are
subtracted from this set to obtain 12,009 highly conserved, gap-free
alignments close to the 5' ends of genes but not within the genes.
Finally, an intersection of the latter set of alignments with results
of query 3 (all SNPs) will produce the set of conserved noncoding
sequences that are close to the 5' ends of genes and also have SNPs.
The result returns 1053 sequences. One of these falls into the 5'
region of the gene encoding SWI/SNF related, matrix associated actin
(SMARCB1/SNF5; Online Mendelian Inheritance in Man (OMIM)
entry *601607; Figs. 2A, 3). Mutations in the coding exons of this gene
are associated with malignant rhabdoid tumors (Versteege et al. 1998 A second example of a complex query is finding CpG islands between closely spaced, divergently transcribed genes; these are good candidates for promoters. Initially, simple queries are used to find genes (not pseudogenes, which are specified under "type" under "Genes") in a forward orientation (+ strand, which returns 16,842 genes), genes in a reverse orientation (– strand, which returns 16,586 genes), and all CpG islands (which returns 26,942 regions). Then on the history page, the results of the simple queries are combined using the proximity feature to identify CpG islands that are within 2000 bp of the 5' ends of genes in the two orientations. This gives 6364 and 5868 CpG islands located close to the 5' ends of genes in forward and reverse orientation, respectively. The intersection of these two sets reveals 1698 ranges containing CpG islands that are within 2000 bp of the 5' ends of divergently transcribed genes. For example, two of the three CpG islands between positions 16,360,000 and 16,410,000 on chromosome 22 are within the set, and examination in the Human Genome Browser illustrates the divergently transcribed genes flanking them (Fig. 2B). Note that one CpG island in this region does not meet the criteria for proximity to divergently transcribed genes. Queries to find CpG islands between divergently transcribed genes can be formulated in a variety of ways. An alternative to the one just described is to find all intergenic regions containing CpG islands, and then find the ones that are within 2000 bp of the 5' ends of genes transcribed in forward direction and also within 2000 bp of the 5' ends of genes transcribed in reverse direction. This returns slightly different results, because the limitation to intergenic regions removed CpG islands in overlapping genes. In addition to the methods just discussed for combining results of simple queries on the history page, we also provide a clustering function. Users can select any number of entries in one category to be within a certain distance of entries in another category. One use of the clustering function is to find combinations of transcription factor binding sites. For example, after obtaining NF-E2 binding sites on chromosome 11 from one query (2555 sites) and GATA1 and GATA2 binding sites from a second query (1906 sites), one can find 15 ranges where the NF-E2 binding site is within 50 bp of two binding sites for the GATA proteins. The proximity function also finds ranges where two different features are within a specified distance, but unlike clustering, one cannot specify the number of features (e.g., binding sites for protein A) that must be close to another type of feature (e.g., binding sites for protein B).
Laj Alignment Viewer Users can easily examine both graphs of alignments and detailed nucleotide-level alignments within the context of annotated genomic features using Laj, as illustrated in Figure 3. This shows the human–mouse alignment located 5' to the human SMARCB1 gene that also contains a single nucleotide polymorphism in humans, which is one of the regions returned by the first example of a complex query. The pip shows the position and percent identity of each gap-free portion of the alignment, and the icons and other tools show where the alignments are relative to the gene and other features of interest. Clicking on a line representing an alignment will display it in the nucleotide-level view (Fig. 3G). This is done interactively, so users can navigate easily along alignments. The information viewed by Laj can be integrated with other information to draw biological inferences. The SNP is a T (highlighted in dark gray in panel G) in the human reference sequence, and it is a G in the alternative allele. The aligning sequence in mouse is also a G. The SNP is within a match to an NF-Y binding site (a CCAAT box) and is just downstream from good matches to binding sites for AP-1 and AML-1 (Fig. 3). Thus, one could hypothesize that this polymorphism affects expression of the SMARCB1 gene by altering the binding affinity of NF-Y (or a related protein) and possibly the interactions of DNA-binding proteins in this region. If an investigator has a strong rationale for studying variable expression of the SMARCB1 gene in human populations, experiments could be designed to test whether the polymorphic site affects expression.
GALA is a relational database that allows users to query across different types of information in multiple locations, find features that are proximal to one another, and store and edit previous queries. Users can investigate an enormous variety of issues using GALA; only a few examples were used in this report to illustrate some of the features. Users may want to look for combinations of features expected for regulatory elements, with a complex query such as "Find all highly conserved noncoding sequences within a specified distance of genes expressed in a certain tissue that also have binding sites for a specified set of transcription factors." Others may be interested in evolutionary issues, such as exploring all ancestral repeats that are conserved between human and mouse and are located close to CpG islands. Others may pursue combinations of clinical (e.g., using the field for "disorders" listed in Online Mendelian Inheritance in Man [Hamosh et al. 2002 ]), biochemical (e.g., using fields for "biological
process" and/or "molecular function" of gene product), and
genetic (e.g., recombination frequency) features. These rich
possibilities should provide new insights that lead to novel
experimentation. GALA complements the existing genome browsers by storing information in a relational database and providing support for complex queries. It uses the browsers both as the source of some of the data and as one of the ways to display the results of queries graphically, including custom tracks for the Human Genome Browser. The data fields used in GALA are gathered from multiple sources in an automated process that allows quick and efficient updates to the database. Many browsers offer similar categories of data, but their current Web servers are not designed to support complex queries across different data fields or different genes, and they are not currently able to return data on the locations of elements that lie in proximity to one another. These features are incorporated in GALA. Enhanced query capacity and integration are continually being added to the genome browsers, such as EnsMart at Ensemble, Table Browser at the Genome Browser, and Map Viewer at NCBI. Complex queries are formulated in two stages. The user initially finds the general categories of interest with simple queries and then uses the combination methods on the history page to filter the results. These methods of combining data include intersections and/or unions of data sets, subtracting features not desired in the final output, and extracting features that lie within the same genomic environment with proximity. In addition to searching for proximity of features, users can find clusters of a designated number of features within a specified range. The proximity and clustering features are flexible tools to integrate multiple kinds of information about genomic regions and their conservation.
We show two sample queries that utilize complex querying and return
data that are relevant to the study of the human genome. The first
query found SNPs in conserved noncoding regions that are proximal to
genes. While these results certainly are not conclusive for a role of
the SNPs in regulating expression of these genes, they provide
candidate sequences that can be examined experimentally. The number of
available SNPs has increased dramatically in the past year, and soon it
may be common to examine their role in regulatory elements. However, at
this time most SNPs are examined as part of the coding regions of
genes. The GALA database could become a primary resource for finding
SNPs in known and candidate regulatory regions. The second query
examines CpG islands that lie between the 5' ends of divergently
transcribed genes. Some of the genes in this category may be in a
genomic environment that contains a bidirectional origin of
replication, a promoter, and a CpG island, such as has been shown for
the TOP1 gene in humans (Keller et al. 2002 Improvements planned for GALA include increasing the data recorded (e.g., alignments from multiple species) and expanding the querying capacity. Currently, queries are limited to those that return no more than 7 million entries, but this limitation should be lifted with planned improvements. Regular updates to new genome assemblies and improved datasets are planned.
Database Management System The database is implemented using the Oracle relational database management system version 8i, running on a Sun Blade 100 workstation under the Solaris 8 operating system. The query form Web pages are created dynamically by CGI scripts written in the Perl programming language, which accesses the database via the DBI Perl module.
Genomic DNA Annotation IDs and accession numbers are used to link the gene entries to information in the resource databases. Data from the resource databases (see below) are automatically entered into the database using programs that parse each file format. We have developed automated procedures to update the data tables, which will be applied to all subsequent updates.
Queries on genes are not limited to the default set of genes; queries
on "alternate gene models" are also supported, such as those
predicted by Ensembl (Hubbard et al. 2002
Information about gene products and their function was imported from
the LocusLink database (Pruitt and Maglott 2001
Information about the tissue-specific expression profile of each gene
with a RefSeq reference is obtained from the UniGene database (Wheeler
et al. 2002
Information relating human genetic disorders to their associated
gene(s) comes from the Online Mendelian Inheritance in Man database,
OMIM (Hamosh et al. 2002
Data for SNPs comes from two sources: dbSNP (Sherry et al. 2001
Information on repeats and CpG islands was imported directly from the
HGB (Kent et al. 2002
Transcription factor binding sites were mapped onto the genome using
the programs Cister (Frith et al. 2001
Recombination data (Hudson et al. 1995
Genomic DNA Alignments
Access
Output Formats Custom tracks for the HGB visualize the results of the query alongside any chosen tracks at the browser. Query results are provided as a clickable button that opens a HGB window slightly larger than the genomic coordinates. Each page is drawn on the fly at HGB and appears in a new window on the browser. Implementation of the custom tracks is described at the URL http://genome.ucsc.edu/goldenPath/help/customTrack.html.
The Java applet Laj is an interactive viewer of alignments and
associated annotations (Wilson et al. 2001 Because GALA provides full annotations for the selected region, it is not practical for Laj to display huge sections of the chromosome at once. The time required for GALA to assemble the data files on very large sections of a chromosome and deliver them across the Internet may cause the connection to time out. Also, the information for annotating very large regions could fill the memory on the user's machine. The time and memory limits can be exceeded even if the query returns only two alignments, if they are too far apart. To avoid this problem, GALA checks the size of the region before trying to display it, and if it is longer than 2 Mb, the program prompts the user to select a subregion for display. These pages appear in separate windows, so that one can examine several such ranges at once. To run the Laj applet, the user's Web browser must support at least Java 1.2, which is now available for most computer platforms (a help page with tips for obtaining suitable Java software is provided). We have deliberately avoided using the latest features from newer versions of Java, in order to retain a wider compatibility. For researchers who wish to use Laj to present their own data, a download package containing the program, installation instructions, and additional documentation is available at our Web site (http://bio.cse.psu.edu/). This package can be used to install Laj as an applet on a Web server, or to run it in "stand-alone" mode to view local data files. The latter mode does not use a Web browser, though it still requires Java 1.2. It provides more capability to compare and manipulate alignments, but cannot visit linked sites.
History Page On the history page, users can see, on average, the last 15 queries submitted, which are saved in a browser cookie for up to one week. Additional queries will overload the allowable cookie size and cause the browser to flush the contents of the history page. Therefore, users should click on the "delete selected queries" button to remove unwanted queries before the cookie exceeds the limit. An option to save queries to a separate local file is also provided. Queries that will be repeated after a lapse of 1 wk should be saved this way to prevent their loss when the cookie expires. Both the cookie and the local file save the choices used to formulate the query, rather than its output. A function to edit previous queries is also available. The old page is opened and a user can simply change the selected fields. Some fields that are in scroll bars should be checked carefully because the default display shows the first entry in the list, even if subsequent entries are selected. The new query will replace the older one on the history page. The history page is designed to narrow the focus of a query. As more data becomes available for each chromosome, queries will require more time. Thus, a feature on the query page allows a user to select the fields of interest and send a place card to the history page, without actually running the query immediately. This option allows a user to specify and combine preliminary queries, without having to wait for the intermediate steps to run. When the final query is submitted all of the searches are performed.
http://genome.ucsc.edu/; The human genome browser. http://www.sanger.ac.uk/; The ensembl genome database. http://www.ebi.ac.uk/; The ensembl genome database. http://www.ncbi.nlm.nih.gov/; Site for Map Viewer and GenBank database. http://www.ncbi.nlm.nih.gov/LocusLink; LocusLink database. http://www.ncbi.nlm.nih.gov/UniGene/; Unigene database. http://www.ncbi.nlm.nih.gov/omim/; Online Mendelian Inheritance In Man database. http://www.ncbi.nlm.nih.gov/SNP/; single nucleotide polymorphism database. http://www.gene.ucl.ac.uk/nomenclature; Human gene nomenclature.
http://sullivan.bu.edu/ http://www.gene-regulation.com/index.html; link to the TRANSFAC database. http://ftp.genome.washington.edu/RM/RepeatMasker.html; RepeatMasker program. http://globin.cse.psu.edu/; homepage for GALA link. http://bio.cse.psu.edu/; homepage for GALA link. http://genome.ucsc.edu/goldenPath/help/customTrack.html; instructions for custom tracks on the human genome browser. http://java.sun.com/products/plugin; download Java Plug-in for use with Laj.
We thank the individuals and institutions providing the publicly available genomic resources used in constructing this database, and W. James Kent for helpful advice. Our work is supported by PHS grants HG02238 (W.M.), DK27635 (R.H) and HG02325 (L.E.) The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 USC section 1734 solely to indicate this fact.
5 Corresponding author.  E-MAIL rch8{at}psu.edu; FAX (814) 863-7024. Article and publication are at http://www.genome.org/cgi/doi/10.1101/gr.603103.
 -like globin genes. Mol. Phylog. Evol. 5: 18-32.[CrossRef][Medline]
Received July 9, 2002; accepted in revised format January 24, 2003. 13:732-741 © by 2003 Cold Spring Harbor Laboratory Press ISSN 1088-9051/03 $5.00  CiteULike  Connotea  Del.icio.us  Digg  Reddit  Technorati What's this?
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
ABSTRACT
RESULTS
mfrith/cister.shtml
