Hybsweeper: a resource for detecting high-density plate gridding coordinates

Cosmid (1), phage, and bacterial artificial chromosome (BAC) libraries (2) have commonly been used for constructing genome libraries, and depending on the organism, this results in thousands of clones that need to be managed. One complication of dealing with such a high number of samples arises during hybridization screening of high-density filter arrays of the genomic clones and trying to distinguish the correct dishes and wells of positive signals. There exists a need for easy determination of clone identities from autoradiographs generated through simple probing methods, overgo probe hybridization (3), and isolation of unique clones from saturation hybridization. To meet this need, we have developed Hybsweeper, a program that is designed to allow quick mapping of positive spots on a filter and straightforward identification of the corresponding clone.

We demonstrate the utility of Hybsweeper by application to an established wheat genome library. The five-times coverage genome library of the 13,500 Mb tetraploid genome of Triticum turgidum subsp. durum is represented by 516,256 BAC clones averaging approximately 130 kb per insert (2). Management of clones for plate replication and spot arraying was provided by the QBot (Genetix, Hampshire, UK), a multifunctional replication robot system. The tested clone library collection was distributed over 1344 384-well microtiter dishes (516,096 clones). As only 48 384-well microtiter dishes could be plated on the QBot apparatus at a time, this collection was enough to create 28 reading sets of the 67,092 Mb of insert DNA, assuming an average of 130 kb/insert.

To facilitate clone retrieval, clones were replicated onto nylon filters where cells were incubated for 14 h, lysed, and prepared for DNA-DNA hybridization experiments (2). The gridding pattern used was one provided by the software included with the QBot system and consisted of 48 384-well plates replicated over six gridding areas (eight 384-well plates per gridding area). Each gridding area consists of a 16×24 array of a 4×4 stamped pattern of clones. Each well of a represented plate was double stamped within the 16 spots using an offset 4×4 stamping pattern. Eight different doublet-patterns were used to allow identification of of the correct plate from among the eight possible. This double-stamping within a 4×4 arrayed area was important in reducing errors in spot picking.

To identify genomic clones that contained genes of interest, several representative gene clones were radioactively labeled and hybridized to the genomic clones on the filters. Filters were either exposed to autoradiographic film or read directly on a radiation signal-detection scanner (Molecular Dynamics Storm™ 820; Amersham Biosciences, Piscataway, NJ, USA). Images were scanned to create either GIF or TIF image formats, typically resulting in a 22×22 cm image with at least a 200 dpi resolution. With the resulting image, the goal was to identify which plate and well corresponded to the positive clone (1 out of 18,432 possibilities). To do this, requires recognizing the doublet pattern (out of eight possible stamped patterns) and then resolving from the pattern and placement the corresponding plate and well.

The software package was written using the Java™ (Sun Microsystems, Foster City, CA, USA) programming language to allow its use in directories on multiple operating system platforms such as Windows® or UNIX-based platforms such as Solaris (Sun Microsystems) or Linux. The program can be accessed through an Internet browser program such as Mozilla Firefox (Mozilla Foundation, Mountainview, CA, USA) or Internet Explorer (Microsoft, Redmond, WA, USA). The image files should be placed in the same directory as the Java hybsweeper.class file and executed by browsing the hybsweeper.html file. Images are then loaded from this environment, and steps are taken to set base coordinates for the image, select positive signal points, and then save the data results.

The most important part of starting the program is to establish the boundaries of the mappable quadrant effectively. This is accomplished by selecting the upper-left and lower-right coordinates. The user clicks the Top Left radio button and selects the corresponding point, then repeats this for the Bottom Right button. After selection of two quadrant coordinates, an outlined frame for the six filter sets and well positions is displayed (Figure 1). The reference coordinates can be moved and checked by alternating between the Top Left or Bottom Right selection mode until the user is pleased with the alignment. Once quadrant coordinates are established, the user then clicks the Pick Spot radio button. As spots are selected, a pair of colored spots are displayed, each representing two of the 16 possible positions for the specific single well location derived from eight different microtiter dishes. For example, the 4–4 and 8–8 positions are horizontal, 2–2 and 7–7 positions are vertical, and the 1–1, 3–3, 5–5, and 6–6 are at diagonal positions. A reference grid pattern is provided on the left-hand side of the display to remind users of the spotting order in each 4×4 array. After each successful selection, the user can click the mouse while holding the CTL button or use the Save Coordinates button to record the value. Once saved, the paired spots change to another color representing a recorded value.

Use of the Hybsweeper program is quite basic and tries to give the user control of point selection while providing helpful cues to make selections as accurate as possible. Additional features to improve functionality include plate series labeling (series of 48 plates at a time), X:Y coordinate position display, spot intensity rating, refresh display, and help.

Spot selection ultimately results in the identification of clones that need to be retrieved. The report can list the recorded spot selections either in the order of which they were picked or in a sorted order by plate. Each report lists as header the image title, the number of clones selected, and the data in tab-delimited rows with the following information: Plate, Row, Column, Score, and Position. When the report is displayed, it can be copied and pasted into a simple text editor. This report may ultimately be used to program colony picking software for the robot to retrieve the selected clones.

With the report displayed, the user has an option to double-check entries. Highlighting a data line and selecting the Show button highlights the spot within the graphical display area, and the button Remove may be selected to remove the data entry from the report.

Reading spot locations manually to resolve plate positions can be quite lengthy, sometimes taking several days, and is subject to human fatigue and error. By using Hybsweeper instead, approximately 300 doublet signals can be resolved in a few minutes time. We have successfully used the software for resolving clone locations in simple probe hybridizations (2), overgo pooling of probes, and probe saturation screening.

Though each point on a filter needs to be selected individually by the user, this can be advantageous when background signals of the autoradiograph cloud some of the hybridization signals. Under these conditions, automated methods can perform poorly. With Hybsweeper, the individual attention per selection allows the technical operator to make a judgment call for each selected spot to assure validity for the selection. While the Hybsweeper software operates as an independent application, it may be possible to integrate it with other packages such as the ImageJ Java package (rsb.info.nih.gov/ij) (4).

Hybsweeper's mechanism for selecting points could also be integrated with a tool such as ComboScreen for multiplex hybridization determinations (5). In cases where several thousand spots are to be selected, an automated strategy such as that described by Audic and Zanetti (6) or available via BioArray (7) or XDotsReader (8) may be desirable or adaptable. However, none of these programs runs on as broad an array of operating systems as Hybsweeper. Complex and automated software, such as the Java-based application High Density Filter Reader (HDFR; Incogene, Clemson, SC, USA), are also available commercially.

The Hybsweeper software package is provided freely for download at wheat.pw.usda.gov/demos/Hybsweeper, and a sample image is provided for testing. The source code for the Java applet is provided and allows for modification. For example, the preferred gridding pattern used by a laboratory may be modified in the software and recompiled using a Java compiler. A discussion list has also been established at the download site for software feedback and suggestions for additional modification.