Guide to GELTRAK


The following guide is reproduced from the "geltrak.doc" file that accompanies these programs.
Geltrak is a system for scanning one-dimensional gels, displaying lane profiles, and calculating areas under the peaks.

Geltrak requires a digitized image in MRC image format. The program tif2mrc included converts most standard tiff format files to MRC image format. (To run the tif2mrc program at the IU MolViz Facility, type tif2mrc your_tiff_file_name). The imedit program is sometimes needed to rotate the image as Geltrak requires lanes to lie horizontally on the screen (To run the imedit program at the IU MolViz Facility, type imedit your_tiff_file_name).

The Geltrak computer program runs on a Unix computer via a colour X-terminal or on a colour graphics workstation. It is interactive and menu-driven and runs in conjunction with an X-windows based library. It displays the digitized image on the screen and allows the user to track along each gel lane. It then displays the profile of the lane from which areas under specified peaks can be calculated. Although the map stores the data in its originally scanned form, the gel is displayed in compressed form to fit on the screen. The compression factors are calculated from parameters stored in the map header. When running the program it is sometimes necessary to expand parts of the profile back to the original resolution in order to use all the available information. Instructions for running the program appear as labels or menus on the screen. The mouse is used for :

To start up the program, type geltrak

The first prompt is for the input filename, e.g., gel.map

It then requests a density range for the map which you must type in; to continue with the default you see on the screen type a / (slash). To make weak bands appear brighter, select 0, some fraction of the upper density, e.g if the range is 1,100000 you might try 1,100. The image is then displayed and the density range can then be re-defined if required. When the map is displayed, you must answer the prompt again; this is to allow you to change the scaling after you have seen the image.

The footprinting option can then be selected; this allows not only two lane profiles to be displayed simultaneously for transferring peak numbers automatically, but also calculates difference probability plots. Select the standard option if you do not wish to do this or 2D analysis if you wish to draw boxes around the areas to be integrated. A postscript dump of the screen can be produced at this stage for laserprinting either immediately or after the program has run.

The user is asked for the left hand averaging window (slit width & height) to be given in screen pixels. This window is displayed on the screen in a position indicated by the user. The right hand averaging window must then be specified; this can be the same as the left hand window or different to provide a variable slit size across the lane. The resulting profile to be drawn on the screen will be averaged from the original input file using the windows just specified. A path along the centre of a lane is requested, and the cursor is used to track this by marking points from left to right along the lane centre. Press the ctrl key and use the left hand mouse button to drag the line to the next point then release the mouse button. Press it again to continue. Release the ctrl key and double click the lh mouse button to finish. Up to 99 lanes can be read in any one run of the program. The lane is displayed with the window bounds above and below.

The baseline is constructed from up to 50 contiguous lines which are indicated in the same way as the lane tracking. There is an option here to construct a the baseline from a single point; just position the cursor and double click the left mouse button. The baseline is subtracted from the profile. The profile will need to be displayed in its expanded form to select peaks for integration. This is done in sections centred immediately above the horizontal point marked with the cursor. The arrow pointing to the baseline marks the last point integrated from the previous re-expansion.

The main menu is then displayed which allows the following options :

  1. Number the peaks in the current lane
  2. Integrate areas by simple addition
  3. Integrate single peaks by profile fitting
  4. Integrate multiple peak sets by matrix decomposition
  5. Integrate areas by 2D Boxes
  6. Switch profile fitting function
  7. Change text font
  8. Calculate/output difference probabilities
  9. Create postscript image dump
  10. Create plotter output file for profile
  11. Track next lane
  12. Recall previously tracked lane
  13. Modify zoom window
  14. Quit

1) Number the peaks in the current lane

This option was principally designed for footprinting gels, but can be used for any gel where the peak numbers need to be associated with particular peaks. It should be carried out before area integration so the the peak areas already have a number associated with them. For standard gels, or in the case where the footprinting option has been selected and the lane is the reference lane, the single lane profile is displayed.

First you must expand the initial area of interest by positioning the cursor and clicking the left hand mouse button. Then start the peak selection by selecting each peak to be numbered. The program calculates the peak maximum automatically and stores this position and the peak number. If it makes a mistake you can correct it by choosing the remove peak option followed by mark shoulder peak option. This option then asks you to mark left and right sides of the peak which it stores instead of the peak maximum. Remember that you can also zoom the area around the peak to see it better.

Where the lane is the comparison lane (bound DNA), both reference and comparison lanes are displayed. Numbers from the reference lane can then be transferred automatically. If you select this menu option, mark the first and last peak positions on the comparison lane. Peak maxima and numbers will be generated and displayed and held in the program for area integration. It is essential to examine this transfer and correct any peaks indicated by the program. These peaks have transferred to the wrong peak and have been moved by the program to a position between two peaks. You may also correct any other peak positions at this stage.

2) Integrate areas by simple addition

This method works for very well-separated peaks and for regions which include many peaks. The user is prompted for an output filename for storing areas. Expand the area of interest next by positioning the cursor and clicking the left hand mouse button. To calculate the area under a peak, mark first left then right positions to enclose the area to be integrated.

Coordinates of the nearest trough in the profile can be read, or absolute cursor coordinates. The right hand vertical line is not included in the sum, but is assumed to belong to the adjacent peak. The option 'start new peak set' means the next position marked will be the left hand minimum beginning the next separate set of peaks. The program works from left to right and will give an error message if you try to go back. Continue with as many regions as desired, re-expanding other parts of the profile as necessary, again always working from left to right. Calculated areas are written to the output file.

3) Integrate single peaks by profile fitting

This method is suitable for integrating areas under well-separated peaks and generally gives a more accurate estimate than simple integration. First select the fitting function to be used. The skew Cauchy has been shown to fit single peaks better than the Gaussian, but some prefer the tried and tested approach.

After typing in an output file name, selecting the fitting function and re-expanding the profile, each peak to be integrated is indicated approximately by the positioning the cursor and clicking the left hand mouse button. The closest peak maximum is located automatically and an approximate fit is displayed over the peak as a series of dashed lines.

A menu appears which allows a least-squares refinement, input of peak start and finish points selected by cursor, a re-estimate of the original parameters, removal of all previous fitted profiles on the screen or conclusion of work on that peak (either including the integrated area in the output file or abandoning the peak). A correlation coefficient is calculated to provide a rough estimate of the goodness-of-fit between the peak and the fitted profile. Areas calculated are written to the output file.

4) Integrate multiple peak sets by Matrix Decomposition

This method is the most accurate for calculating areas from overlapping peaks but is only appropriate for lanes with many peaks. It works by fitting each peak to a function, either skew Cauchy or Gaussian and then decomposing the peaks to calculate their individual areas.

The output file is specified then the fitting function is selected (skew Cauchy or Gaussian). Three ways of peak fitting are offered :

i) Calculate regression coefficients by selection/
Type in regression coefficients/
Read a file of regression coefficients
Use this method if you have a few well-separated, regularly-shaped peaks and many overlapping peaks.

This method fits a small number of well-separated peaks to the selected function by Least Squares, refining width, height, and position (shape and skew are also refined for the skew Cauchy). It uses parameters derived from these fitted peaks to calculate parameters for the overlapping peaks.

To do this, re-expand a section of the profile containing some well-shaped single peaks. It is worth noting here that it is better to use that half of the gel furthest from the loading point as peaks close to the origin are frequently multiple peaks. Working from left to right, use the cursor to indicate each peak which can be profile-fitted and refined.

The fitted peak is displayed on the screen as a dashed line superimposed over the original peak. All the peaks accepted by the user are used to carry out an Analysis of Variance which calculates a regression line to represents width/height ratios (and for the skew Cauchy a second regression line to determine shape/height ratios) varying with distance. If the regression is insignificant at 5%, the value furthest from the line is rejected and the procedure repeated until convergence. The regression coefficients calculated here in one lane can be saved and applied to other lanes as long as the lanes are comparable.

ii) Calculate width/height ratio automatically
Use this method if you do not have at least 6 decent well-separated peaks. The results will not be quite as good and consistency is important, so you should treat all lanes in the same way. Regression coefficients are calculated over the entire peak set.
iii) Calculate individual ratios automatically
Use this method if the others fail or if the peaks widths do not vary with distance. It does not use a regression line but fits each peak individually. The profile overlaid with the fitted peaks can be displayed and the user can refit and re-refine any of the peaks individually.
Methods ii) and iii) sometimes require user intervention for very poorly shaped peaks - it will ask you to mark left and right sides of such a peak Peaks are rejected automatically if they are too small to fit.

Where the standard option was selected at the beginning of the program, the set of peaks to be decomposed into component parts is selected by using the cursor at each approximate peak maximum. If peaks have been previously numbered,or if the footprinting option was selected, a range of peak numbers is typed in. e.g. 1,23 or 15,35 or the whole range can be selected. Peaks are fitted by one of the three methods already described, using the Levenburg - Marquardt technique.

The entire profile can be expanded and displayed with fitted peaks overlaid before starting the decomposition stage. Decomposition results are displayed on the screen and written into the output file. Small peaks are sometimes buried by the tails of larger neighbours. When this happens, the peak areas are returned as negative numbers. Geltrak removes these peaks and recalculates the decomposition. The user can accept the decomposition or try one of the other profile fitting methods. The areas calculated and peak distances from the origin are written to the output file.

5) Integrate areas by 2D Boxes

This option first gives the option for background sampling which should be made by selecting four distinct areas in appropriate areas of the image. Select bottom left, then top right corners of the box to be integrated. Areas are written to the output file.

6) Switch profile fitting function

This option allows the fitting function to be switched from Gaussian to skew Cauchy or vice versa. Note that the regression line may need to be re-specified after this option if calculation by matrix decomposition is to be performed.

7) Change text font

This option allows the user to change the font of the numbers displayed on the screen.

8) Calculate/output difference probabilities

To use this option, you must have used the footprinting option and tracked, numbered and integrated a reference lane and at least one comparison lane. An output file name is specified, then select the divisor. Bound dna lane numbers are entered successively until all the comparison lanes have been typed in. Probabilities and log differences are then written to the output file.

9) Create postcript image dump

Grey-scale or colour postscript files can be generated and are always called Ximage.ps

10) Create plotter output file for profile

Output files can be created from each profile which can be plotted on the printer using a program called curvy. File names, titles, horizontal and vertical scaling are prompted for at this stage. A menu gives a choice of output to line printer, laserprinter or saving the file for any other purposes.

11) Track next lane

This moves back to window size selection before tracking the next lane.

12) Recall previously tracked lane

This option prompts for the lane number to be recalled. The baseline-subtracted profile for this lane is displayed, and it then becomes the current lane.

13) Modify zoom area

This option allows the user to change the size of the zoom area and the magnification area.

14) Quit

Quit.


Back to  |  Application Guide  |  MolViz Home  |
Send comments to chemvis@indiana.edu
Last updated: 01/23/2001