FPC Web Tools for Rice, Maize, and Distribution

Rice Map

The rice FPC map has 72,703 clones, 8,870 markers, 180 contigs, and 2,918 anchors. FPC calls any marker that has a location on a chromosome or linkage group an anchor. There are two types of anchors: (1) frameworks are well ordered and (2) placements are binned between frameworks. For the rice map, the 1,378 frameworks are the Japanese Genetic markers (Harushima et al., 1998). The 1,040 placement markers have the prefix OJ and originate from Monsanto-International Rice Genome Sequencing Project Map Integration (Chen et al., 2001). Chen et al. (2001) reported 458 contigs, which have since been reduced to 180 (H.R. Kim, personal communication). Of the 180 contigs, 93 are anchored to chromosomes by the genetic markers. Of the remaining contigs, 49 have only two clones and 26 have less than 10 clones; these contigs probably contain clones with bad fingerprints and are generally ignored. By these criteria, only 12 good contigs (i.e. with ≥10 clones) remain unanchored.

The initial WebFPC display for the rice physical map is shown in Figure 1. It provides options to search by clone, marker, or contig. If a marker is contained in more than one contig, all contigs containing that marker will be listed. Substrings can be used for marker and clone names. For example, to view all the sequenced clones, one would enter the string “sd1” (the “sd” stands for simulated digest, as explained below).

WebFPC is implemented in Java, which allows fast navigation around an entire contig, avoiding the slow redisplay common in Web displays using the paging method. Each marker is centered over the largest stack of clones to which it is attached. WebFPC features a filtering window that gives user options to show or hide information. For example, there may be many markers in a small region, causing the marker track to become very deep. Marker filtering allows the user to limit the depth by showing only the markers of interest, as shown in Figure 2a. In Figure 2b, markers prefixed by SOG or OJ have been removed from the display. Anchors are shown at the bottom of the display (Fig. 2c). By default, only the frameworks are shown, but the placements can be made visible via the filtering window. While only a region of the contig may be visible, all frameworks are shown, providing an overview of the whole contig. Selecting an anchor centers the contig display on the corresponding region. A pull-down at the top of the display lets the user filter the clones by “No Buried,” “All,” or “Seq Only.” A buried clone is one whose fingerprint pattern matches that of another clone either exactly or approximately. Selecting No Buried hides the buried clones to limit redundant information. The Seq Only option shows only the simulated digest (SD) clones, which are generated from sequenced clones.

The SD clones are generated by a nightly cronjob (a script that is scheduled to run automatically at a given time). The cronjob executes the following steps: (1) download updated rice sequence from GenBank; (2) run a simulated digest on each sequence (Engler et al., 2003) to create an in silico fingerprint; (3) if a sequence digests into more than 55 bands, breaks it up into multiple overlapping clones, where they are consecutively suffixed by sd1, sd2, etc. (this is done because a clone with many more bands than the typical clone will have a low probability of overlap with the other clones); (4) add the fingerprints to the FPC database; (5) compute the overlap probability between each SD clone and all other clones in the database; (6) automatically place each clone in the same position as its best match; (7) generate a file of links between FPC SD clones and GenBank sequences; and (8) update the WebFPC site. The number of SD clones are shown on the initial display (Fig. 1), and they are also colored yellow in the contig display (Fig. 2b) and remarked (Fig. 2c). The remark is the original clone name, the author, and the chromosome assignment. Since this information is automatically parsed from the GenBank record, it may not be correct if it was not entered into GenBank in the canonical format. But the automatic entry means that WebFPC is always up to date with the latest sequenced clones. The rice sequencing status page is also updated nightly and shows the coincidence score between the original fingerprinted clone and the SD clone; this provides both clone and sequence assembly confirmation (see www.genome.arizona.edu/shotgun/rice/status).

In the top right hand corner of the contig display, there is a pull-down that lists other on-line databases to which the user can link. For the rice map, there are links to GenBank (Benson et al., 2004), Gramene (Ware et al., 2002), and RGP INE (Harushima et al., 1998). Since RGP INE is selected in Figure 2, all markers and clones with links to records in RGP INE are highlighted. Selecting a marker and then selecting “Link to Site” will bring up the corresponding record in the given on-line database.

Figure 3 shows the rice WebChrom display, which provides a view of the ordering of the contigs and anchors along the chromosome. Selecting a contig will bring up the WebFPC display for the contig. The markers link to WebFPC, Gramene, and RGP INE. It is not unusual for the genetic location of anchors to disagree with their location in a contig. For example, contig 14 in Figure 3 shows all of its markers in close succession (bounded box with tick marks representing genetic markers), but the long unbounded yellow box indicates that one marker is on the lower part of chromosome 3. The contig's chromosome and the position on the chromosome are calculated by FPC (Engler et al., 2003). A contig is assigned to a chromosome based on a majority rule of all the anchors in the contig. The contig's position is calculated from all anchors located on the assigned chromosome, excluding anchors that are too far from the average location.

If the user wants to search for a particular marker on the chromosome map or see the distribution of a set of markers based on a substring of the marker name or marker remark, he or she can use the WebChrom Search tool. Many markers are from expressed sequence tags, making it advantageous to remark them with their annotation. As a test case, we blasted (Altschul et al., 1997) all the markers against SwissProt (Boeckmann et al., 2003) and added the highest scoring hit as a remark for each marker. Since the remark contains the word SwissProt, the user can search on it to view the distribution of annotated entries (see Fig. 4).

Though WebFPC shows overlapping clones, the amount of overlap is not exact due to the error in the data (Soderlund et al., 2000). Hence, if a user wants to know which clones strongly overlap with a given clone, he or she can do so by using the WebFCmp tool shown in Figure 5. The user inputs one or more clones and a cutoff. An input clone is compared with each clone in the FPC database by first counting the number of bands N that have the same value within a tolerance, then computing the probability score that the N bands are shared by coincidence (Sulston et al., 1988; Soderlund et al., 1997). If two clones have a score below the cutoff they are said to overlap. WebFCmp outputs all overlapping clones with links to the corresponding contigs in WebFPC.

Suppose one wants to determine whether a sequence from a related organism is found in rice and, if so, what other markers are surrounding the given marker. Such information can be gained by querying all sequences associated with FPC clones, namely the sequences for all SD clones and/or the bacterial artificial chromosome end sequence (BES) for all fingerprinted clones. FPC has a feature called Blast Some Sequence (BSS) that blasts a file of sequences against a directory of BESs or genomic sequences and creates a report of all hits and their location on the FPC contigs. For the WebBSS, the user inputs a sequence, the FPC BSS routine is called, and the output is parsed and displayed as shown in Figure 6. Selecting a contig will bring it up in WebFPC, where the user can view the surrounding markers and clones. As of February 7, 2005, there are 4,058 genomic sequences and 98,286 BESs.

For an alternative view to WebFPC and WebChrom, we have developed an FPC configuration file and GFF file description that are used to create a Generic Model Organism Project (GMOD) genome browser (Stein et al., 2002). This allows the user to view all the contigs and frameworks for a chromosome in a bird-eye's view. Selecting a region of genetic marker displays the corresponding region of the contig in the bottom section. Browsing can be done by chromosome or by contig. In chromosome browsing, contigs are aligned to a chromosome based on the position calculated by FPC. The length of a contig is the number of CBunits (i.e. approximately the number of consensus bands in that contig). The approximate length in base pairs is the number of CBunits times the average band length (typically around 4,096 bp for agarose). The length of the chromosome represents the total length of all its contigs with 100 bp between contigs. Features like markers, frameworks, clones, and sequenced SD clones are shown in different tracks (see Fig. 7).

A full description of the origin of the clones and markers in the rice FPC is presented at the Web site (www.genome.arizona.edu/fpc/rice). The FPC file is available from our FTP site (ftp.genome.arizona.edu/pub/fpc/rice/).

Maize Map

The maize genome (approximately 2,500 Mb) will be the next cereal genome to be sequenced and will thus require a robust physical map and software tools to support the sequencing project as well as a variety of positional cloning projects. The maize FPC map currently contains 292,168 clones, 17,523 markers, and 2,998 anchors (Coe et al., 2002; Cone et al., 2002). A total of 13,810 of the markers are overgos derived from expressed sequence tags (Gardiner et al., 2004). The anchors are from the IBM neighbors map, of which the 1,931 frameworks are the IBM genetic markers and the 1,067 placements are from other genetic maps that have been binned in the IBM map (Fang et al., 2003). There are currently 766 contigs, of which 413 are assigned to chromosomes (release October 27, 2004). The maize FPC map is still under construction. The set of markers will continue to change as new ones are added and faulty ones are removed. The contigs will continue to change as they are being manually edited to merge contigs and split chimeric contigs (F. Wei, personal communication). The contigs are located on chromosomes using the FPC routine, but due to false positives, false negatives, and ancient polyploidization (Paterson et al., 2004), there is a fair amount of ambiguity. The locations are often manually edited so they are positioned correctly on the chromosome. For major releases, the contigs are renumbered to reflect the ordering along the chromosome, which gives the contig number a meaning in relation to the other contigs.

The maize FPC Web site has the same set of tools as the rice Web site. WebFPC links to maizeGDB (Lawrence et al., 2004), iMap (Fang et al., 2003), and GenBank (Benson et al., 2004). Both maizeGDB and iMap link to WebFPC based on a marker or clone name. WebChrom links to GenBank, maizeGDB, and WebFPC. The maize contigs tend to have many long yellow unbounded boxes indicating the ambiguity of marker locations on the contigs.

For the WebBSS, there are currently (as of February 7, 2005) 490 genomic sequences and 682,116 BESs to search against. As with rice, new and updated sequences are downloaded nightly so all available sequences are in the database. The maize FPC map currently contains 503 SD clones, of which 13 have the suffix of sd2 or greater, indicating they come from GenBank sequences that result in more than 55 bands. The number of SD clones will continue to grow since clones are currently being sequenced. The maize sequencing status page shows what clones have been sequenced and where they are located, along with their similarity to the original clone (see www.genome.arizona.edu/shotgun/maize/status).

A full description of the origin of the clones and markers in the maize FPC is presented at the Web site www.genome.arizona.edu/fpc/maize. Additionally, a High Information Contig Fingerprinting (HICF; Ding et al., 1999; Meyers et al., 2004; Nelson and Soderlund, 2005) map has been assembled by FPC and can be viewed from this URL.

Testing the Large Numbers of Anchors in the Human Map

Rice has a finished map and an almost finished sequence. Maize has almost 300,000 clones, and the map and sequencing are still in progress. These two datasets provide good test cases for maps with a large number of sequences and clones, respectively. To test the tools on a dataset with a large number of markers, we downloaded the human map from www.bcgsc.bc.ca/perl/humanbac (The International Human Genome Mapping Consortium, 2001), which has 69,507 markers and 26,164 anchors. WebFPC and WebChrom were installed for the human map (www.agcol.arizona.edu/fpc/human). The only noticeable increase in wait time is during the display of results for the WebChrom Search software.

Setup and Distribution

The FPC tools with the prefix of Web are distributed as a package. A difficulty in distributing code for the Web is that it is a mix of Java, HTML, and CGI, where the three types of files go in different directories. An additional complexity is that the WebAGCoL package is composed of four tools with different requirements. A manual could be written to explain how to set up the files, but that would be tedious and error prone. Hence, we have written a setup script that reads a configuration file and automatically runs the correct scripts and puts files in their correct directories. It also creates a script that can be run to update the Web sites based on an updated FPC database.

WebAGCoL Package

The WebAGCoL package contains WebFPC, WebChrom, WebBSS, and WebFCmp, along with a setup script and on-line help. Each tool has a preprocessor script that creates the files necessary for fast display. They all read a shared configuration file. All preprocessor scripts expect to read an FPC file written with FPC V7 or a later version. All preprocessors are written in Perl, all graphics except WebFPC are generated using the GD library (www.boutell.com/gd/), and Perl CGI is used for run time execution. At AGCoL, all processing and updates are performed on a Sun 280R with 2GB RAM.

WebFPC is written in the Java programming language and is implemented as an applet. The preprocessor perl script splits the FPC database at a contig level, writing information for each contig as XML. This allows a Java SAX (www.saxproject.org) parser to retrieve and parse information simultaneously, reducing the time spent waiting for a contig to be displayed. The preprocessor reads a site specific file to determine how to color clones and markers. For example, for the rice (Oryza sativa) and maize (Zea mays) FPCs, the sequenced clones and electronic markers are highlighted yellow. It reads the directory of reference files to set up links to external sites. The preprocessor also creates the initial HTML that accesses the WebFPC Java Jar file. A CGI script is used to execute an external lookup in order to display a contig without going through the initial page.

The WebChrom preprocessor splits the FPC file into one HTML file per chromosome and writes perl GD code to display the graphics. The WebChrom Search tool stores the markers in a “Storable,” which is a hash table and is used by the CGI script for fast searching.

For WebBSS, the preprocessor creates a modified FPC file that contains only the necessary information; this speeds up the reading of the file during the Web-based execution. The configuration file contains the paths to the BES directory and genomic sequence directory. FPC is run in batch mode to execute the BSS. The BSS saves the output in a file, which is read by a CGI script and displayed.

For WebFCmp, the preprocessor creates a modified FPC file that only contains the index into the file of bands in order to increase speed. FPC is run in batch mode to compare the fingerprints. A CGI script is run to read these files, compute the overlap score, and display the results.

The setup script reads a configuration file to determine where to put the CGI, HTML, and Jar file. It also reads the location of the reference file, FPC file, and target directories and runs the preprocessors. It creates the initial HTML file and writes a file called update.sh that can be used to update all the Web tools when a new version of the organism's FPC file is updated.

Processing at AGCoL

For the rice and maize Web sites, a cronjob is run nightly to download new and updated sequences from GenBank (as mentioned previously). Each GenBank file is parsed into a FASTA file of sequences and saved into a directory whose location is known by WebBSS; hence, this feature is always run on the latest sequences. A program called FSD2 (FPC Simulated Digest, Version 2) reads the GenBank file, cuts the sequence into overlapping clones, and creates the file of restriction fragment sizes. It also extracts the first author, clone name, and chromosome and writes the information into a file to be loaded as a remark for the clone. The size2band program uses the file of marker lanes (used by Image) to convert the sizes to bands. FPC is then run in batch mode to enter the new fingerprints and remarks into the database and position each clone in the same location as its best match. Once a clone has been positioned in FPC, it is not automatically repositioned when an updated record is entered. Therefore, we periodically remove all the SD clones from their contigs by making a keyset of them and executing “Move to Ctg0” from the pull down menu. They are then repositioned by executing “Keyset->FPC” on the keyset of SD clones. FSD2 is available from the FPC Web site.

BioPerl FPC Module

BioPerl (www.bioperl.org) is an initiative that seeks to simplify bioinformatics development by providing perl objects that perform mundane tasks such as parsing a file and retrieving information from it. To further this undertaking, we have developed a BioPerl module that reads an FPC file and allows the user to extract information from it. For example, one may retrieve all markers in a particular contig or find all clones attached to a marker. This module also converts FPC data into GFF format suitable for input into a Generic Genome Browser database, discussed in the next section.

GBrowse for FPC Map

The GMOD is “a joint effort …to develop reusable components suitable for creating new community databases of biology” (Stein et al., 2002; www.gmod.org). This community has developed GBrowse, which is a viewer designed to present linear genomic data on the Web. A GFF file is created to populate a mySQL database, and a configuration file is created that describes the tracks to be displayed in the GBrowse. The GBrowse software reads the configuration file and extracts data from the database in order to display the data. We have created the configuration file for displaying the FPC data. The BioPerl FPC module creates the GFF file to be loaded into the database.

Web Resources

• www.agcol.arizona.edu/software/fpc: (1) FPC software, (2) tutorials, (3) links to other Web based FPC displays, and (4) the WebAGCoL, FPC GBrowse, and FPC BioPerl.

• www.genome.arizona.edu/fpc/rice: The Rice FPC.

• www.genome.arizona.edu/fpc/maize: The Maize FPC.

• www.genome.arizona.edu/fpc_hicf/maize: The Maize HICF FPC.

• www.agcol.arizona.edu/fpc/human: The Human FPC.

• www.genome.arizona.edu/shotgun/maize/status: Maize sequencing status.

• www.genome.arizona.edu/shogtun/rice/status: Rice sequencing status.

• www.bcgsc.bc.ca/perl/humanbac: The human FPC download.

• www.maizegdb.org: A central repository for public maize information.

• www.gramene.org: A curated comparative genome analysis in the grasses.

• www.maizemap.org: Maize mapping project.

• www.maizemap.org/iMapDB/iMap.html: iMap display of maize genetic markers.

• rgp.dna.affrc.go.jp/publicdata/geneticmap98/geneticmap98.html: Rice genetic markers.

• www.bioperl.org: Open source Perl tools for bioinformatics, genomics and life science research.

• www.gmod.org: The model organism system databases.

• www.ebi.ac.uk/swissprot: Protein database of annotated protein sequence.

• www.realvnc.com: VNC a tool to view and fully interact with one computer from any other computer.

• www.boutell.com/gd/: C library for the dynamic creation of images.

• stein.cshl.org/WWW/software/GD/: Perl Interface to GD Graphics Library

• www.saxproject.org: Java API for XML.