Skip to content

Encapsulates Astromatic' Sextractor, a software for objects detection in astronomical images.

Notifications You must be signed in to change notification settings

dockerstuff/docker-sextractor

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

30 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Introduction

This project from the Minden lab is a way to help you apply Source Extractor to detect changes in DIGE gels.

Source Extractor is a neural-network based star/galaxy classifier that we've also found to be useful for identifying and quantifying protein spots in DIGE gels. The advantage is that it’s free and open source, so we know where the values are coming from, which provides a more accurate and transparent way to measure protein changes. Here’s a 3-minute video overview of the project.

You can check out the on-line documentation, the official web page, and the user forum.

Here’s a guide to take you from TIFF files to protein changes:

1. Install Docker and SourceExtractor

  1. Docker is a way to install and run software on many platforms, similar to a virtual machine. This makes installing Source Extractor much easier and more secure. Download it here.

  2. Github is a way to host and update files. Download docker-sextractor from our github.

  3. Navigate to the Github folder you downloaded, and double click on: run.command

  4. Optional: Install some useful DIGE and SExtractor macros:

Copy DIGE_SExtractor_tools.txt from ImageJ Macros to ImageJ/macros/toolsets

2. What's included

  • run.command : Clickable script that handles installation/running Source Extractor.
  • io : Folder that contains all three input folders (input_fits_3, input_fits_5, and input_fits_sum) and will be the location of output folders, with timestamps.
  • config : Folder that contains parameter files (DIGE.param and DIGE.sex), and gauss_2.0_3x3.conv, a convolution mask which might need to be edited in advanced cases.
  • ImageJ Macros : Folder that contains DIGE_SExtractor_tools.txt, which is a file containing several helpful ImageJ macros for vizualizing SExtractor outputs, as well as an example output image.
  • backend : Contains scripts that manage input/output files while using Source Extractor. You shouldn't ever need to edit or run these.

3. Prepare files

  • In ImageJ, save the Cy3 and Cy5 TIFF images as FITS files in the folders input_fits_3 and input_fits_5

  • Make a merged sum: paste control → add, copy and paste one window on to the other. Save the result as a FITS file in the folder input_fits_sum

  • Make sure all files have the same name

  • Click on run.command

Afterwards you should find:

  • Aper files are masks that you can open in ImageJ to see if the spots it identified are reasonable. Dotted lines mean less confidence in identifying that spot. These files are found in output_image_3 and output_image_5

  • output_numbers_3 and output_numbers_5 hold csv files that contain the raw intensity values. You can open these in excel and make the columns index, x, y, raw intensity.

If the y-coordinate is inverted, in ImageJ check analyze / set measurements / invert y coordinate

4. Tune segmentation parameters

Coming soon to a readme near you!

This section should include: Key parameters, Intuition on what they mean, How to adjust them, Recommended range, Possible example images

ANALYSIS_THRESH 1.7 Threshold to start running analysis, measured in in number of sigmas over background RMS.

DETECT_MINAREA 10 minimum number of pixels above threshold

DETECT_MAXAREA Can be used to avoid objects larger than a set number of pixels, for example, Drosophila yolk.

DETECT_THRESH 1.7 <sigmas> or <threshold>,<ZP> in mag.arcsec-2

DEBLEND_NTHRESH 64 Number of deblending sub-thresholds

DEBLEND_MINCONT 0.00001 Minimum contrast parameter for deblending 0 picks up the faintest objects/ 1 turns deblending off. For faint images, we recommend 0, otherwise turn up to 0.0001

BACK_SIZE 16 Size of the background mesh: <size> or <width>,<height> The default is 32, but I recommend 8-16.

If this is too small, the background estimation gets distracted by objects and noise, and the signal gets absorbed into noise. If this is too big, you miss the small-scale variation in background.

32-256 is normal for stars. I thought DIGE would have much less background variation, meaning bigger is better, but 16 seems to be a good compromise based on fly data. 8 is also reasonable, and will give smaller ROIs.

BACK_FILTERSIZE 3 Background filter: <size> or <width>,<height>

BACK_TYPE MANUAL 

still trying this out

BACK_VALUE 300

still trying this out

What's ASSOC?

5. Decide on guidestars

How do you look at raw values in excel?

You can visualize these coordinates in ImageJ/Fiji:

  • Install the ImageJ macros as described in 1.4 above
  • In ImageJ, click the double red arrow on the right and load DIGE_SExtractor_tools
  • Click the button labeled “Import XY… Tool”
  • Select your CSV file You should get something that looks like “example of imported coordinates.png”

If you want to only load some points, you might want to copy and paste only those coordinates into a new CSV file. Any columns that you delete in excel you also have to select and "clear contents" or there will be extra spaces and the macro will get confused.

  • Across at least 3 biological replicates, look at all the raw values. Find 5-6 spots that are evenly distributed through the gel (the macro above helps) that don't change more than 6% between Cy5 and Cy3.

  • Take a ratio of Cy5 to Cy3 for each guide star, take an average, and multiply the raw values by that correction factor.

Consider making the ratio less than 1 to prevent stack overflow. A good correction factor is within 30%.

6. Quantify changes

  • In Excel, multiply one channel by the correction factor. Now all the guide stars should be about the same between channels.

  • Then calculate fold ratio changes, and set a reasonable threshold.

More than a hundred changes is probably too many.

Rejoice! You’ve identified and quantified protein changes in DIGE gels!

About the authors

Ian and Ardon are pretty cool.

Troubleshooting importing XY coordinates into ImageJ

It's possible to import points from a text file too, provided you put it in this format:

	points
	n
	x1 y1
	x2 y2
	...
	xn yn

Where n is the total number of points you have.

For troubleshooting purposes, make some dummy points and use the Export XY macro to see the format it expects. You can then copy your data over that file.

Note that the Import XY macro inverts the Y coordinate in line 41:

ypoints[i-1] = parseInt(1023-line[iY]);

If you want to keep the Y coordinates as they are, just delete the 1023- part.

About

Encapsulates Astromatic' Sextractor, a software for objects detection in astronomical images.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •  

Languages