STAC

This document describes the use of STAC Station, an application for performing STAC analysis and for viewing the results of STAC analysis. See also the STAC tutorial. STAC analysis requires aberration data for each of a number of different samples and it performs a statistical test to determine locations which are concordantly aberrant across individuals more than should be expected by chance. The statistical theory and algorithm are desribed in the STAC technical manual.

Reference

Diskin SJ, Eck T, Greshock J, Mosse YP, Naylor T, Stoeckert CJ Jr., Weber BL, Maris JM and Grant GR. (2006) STAC: A method for testing the significance of DNA copy-number aberrations across multiple array-CGH experiments. Genome Research (in press).

Installation

• Mac:
   
  Create a directory somewhere on your system where STAC Station will live. Download STACStation_1.1.tar and save it to the directory you created. Open a terminal shell and move into the directory. Unpack with the following command:
  
  > tar -xvf STACStation_v1.1.tar
  
  To execute STAC enter the command:
  
  > java -Xmx256M -jar StacStation_v1.1.jar
  
  The STAC window should open. If you do not have 256 Mb of memory availalbe this might not work, in which case try changing 256 to 128. It's not advisable to lower it below 128 as you might run out of memory while trying to run it.
   
  
• Windows:
   
  Download STACStation_v1.1.zip and save it to the directory where STAC Station is to live. Unzip the archive STACStation_v1.1.zip to the directory.
   
  To execute STAC double click on the StacStation_v1.1.jar file. Alternatively, STAC Station can be started as follows:
   
  Open a command prompt (this should be under Start  Program Files  Accessories  Command Prompt). In the command prompt window move into the directory that contains the STAC file. You can do this by typing "cd" followed by a space, and then drag-and-drop the directory into the command prompt window and hit enter.
   
  enter the command:
  
  > java -Xmx256M -jar StacStation_v1.1.jar
  

1. Location data:
   • The stretch of DNA is represented by a sequence of locations (e.g. 1Mb, 2Mb, 3Mb, ..., 100Mb).
   • The data consists of a (binary) aberration call for each location of each sample, represented by a 0 (no aberration) or a 1 (aberration).
   • There must be a tab-delimited header row followed by tab-delimited rows of aberration data, one row per sample.
   • Example Location Formatted Input
   • Read more about pre-processing your data to be location formatted.
2. Span data:
   • The first line of the file must give the span of the entire chromosome arm, or the span of whatever region is to be anlayzed.
   • Each consecutive line gives the spans for each experiment, separated by semi-colons
   • Example Span Formatted Input

Go here for more notes on formatting and pre-processing

Usage

1. "Open to Run" which takes a STAC input file to be analyzed, as described above.
2. "Open to View" which takes a STAC analysis results file and opens it just to view.

Note: STAC Station can open one file at a time to analyze, or a directory of files. Once a file or directory is open to analyze, the "ANALYZE" button becomes enabled.

STAC can also view multiple files (the number of files you can open is limited only by memory - if you run out of memory change the "256" in the command used to execute STAC Station to something higher). The file opened to analyze does not appear in the results viewer until "ANALYZE" has been pressed and the analysis is complete.

As you perform analyses on files, they get added to the viewer list. Similarly as you open results files to view, they also get added to the list. You can then browse through the results files using the forward and back arrow buttons. The current file number and the total number of files is shown in the status bar at the bottom of the window.

Analysis Options

• STAC is based on a permutation test. This sets the number of permutations to use. The default is 100. The run time can vary tremendously depending on the data set, so in some cases 100 permutations will take seconds and in other cases it could take much longer. If your data set is running very quickly you might want to raise the number of permutations to get p-values accurate to more decimal places. Going from 100 to 1000 can make some small difference and give less variable results. Going to even more permutations, for example to 10,000, shouldn't make much difference. There is probably no point in going over 10,000.
• STAC uses a herustic approximation, the accuracy of which depends on something called the search parameter. The default is 3,500. The higher this parameter is set the more powerful the method will be. However the default gives generally good results.

Display Options

• Confidences. STAC uses two statistics, the "frequency" and the "footprint". In most cases the footprint will be more powerful at all locations. But sometimes the frequency is better, so that option is available. STAC Station does not display p-values, but rather displays "confidences" which are one minus the p-values. The footprint and frequency confidences can be displayed separately or together. Furthermore, each can be displayed with bars or with a line. To turn them on or off use the four confidence buttons on the task bar. These can also be turned on using the "Options  Display" menu.
• Frequencies. You can turn on the frequency line using the "Options" in the pull down menu. Note the difference between the frequency line and the frequency confidence line. The frequency line simply gives the frequency of aberration for each location. The frequency confidence line gives the actual confidences of the frequencies. If a row of data contains no aberrations then it is omitted from the view, however you can still give the frequencies based on all the data using the "global frequency line" option. If this option is to be used, make sure to include also the rows in the input file which consist of all 0's.
• Colors. The colors of the frequency line, the global frequency line, the frequency confidence and the global frequency confidence can be set using the menu Options  Display  Colors.
• Highlighting Positions. Using the Options  Display   Enter Positions to Highlight menu item you can select a position for which the intervals that contribute to the confidence will be highlighted. Enter the position using the number of the position given in parentheses.
• Font Size. You can increase the experiment ID and positions font independently using the buttons on the task bar. You can also fit the view to the screen.

Output

• Save to jpeg - you can save the current file being viewed to jpeg, or all open files to jpeg.
• Reports - you can export the analyses as report files which give the frequency and footprint p-values for all locations in tab delimited text format.
• Note: STAC analyses are automatically saved to the disk to view again later without having to rerun the analysis - if you opened a STAC input file and ran STAC analysis, the results file is saved under the same name as the STAC input file with ".stac" appended to the end.

General Operating Characteristics

• Sample size: In general,increasing sample size will increase the power of STAC to detect significant concordant aberrations at a cost in run time (dependent upon the number of intervals per sample).
• Location width: In general, decreasing the fixed-width location size (see notes on formatting and pre-processing) will allow for finer-resolution mapping of significant concordant aberrations. This should not have a significant impact on run time, however we generally recommend setting the location width approximately equal to the resolution of the array.
• Number of probes: Increasing the number of probes in a region will allow for finer-resolution mapping of concordant aberrations and this should be reflected in the use of a smaller fixed-width location size used in the input. Increasing the number of probes alone will not increase the run time of STAC. However, if an inappropriate method for calling aberrations in individual samples used (e.g. using thresholds for a SNP array), then you may see a significant increase due to fragmented intervals in the input.
• Number of intervals: Data sets with fewer intervals of aberration will run faster than those with many intervals. In extreme cases (exceptionally large number of intervals compared to the current "average" data set) execution time can be affected significantly. We are currently evalutating approaches to better accomodate these extreme situations.
• Number of permutations: to a point, increasing the number of permutations will decrease variability seen in the results. We have seen differences between 100 and 1000 permutations, far less between 1000 and 10,000 permutations, and no differences ever after 10,000 permutations. Increasing the number of permutations will increase the runtime; the amount of this increase is dependent upon sample size and then number of intervals as well as the value of the search parameter.
• Search parameter: In general, increasing the search parameter will increase the power of the heuristic search implemented in STAC with a slight increase in run-time. However, for very large data sets containing many samples with many aberrant intervals, increasing the search parameter significantly (say to 10,000 or more) will increase the run-time substantially but may produce the same results as obtained with a lower search parameter. We have found that the default produces good results on the average cancer data set containing 50 samples with a high frequency of aberration.

  NOTE: If STAC is taking a very long time to run on your data, please contact us at diskin@email.chop.edu to obtain a distributed version (STAC-Grid); we will be glad to help you install and configure this. STAC-grid parallelizes the permutation portion of the algorithm and in most cases can greatly speed up execution time.