Quantification of the fluorescence intensity in diffraction limited spots, imaged by fluorescence microscopy.
1.04 - The threshold value used to compute the mask with the Yen algorithm is now computed on the image z max projection. That way the threshold values are smaller and the spot selection less stringent. Otherwise, dim spots were often discarded.
The software is distributed under the GNU General Public License v3.0. However, we would appreciate if you cite Picco, A., Kaksonen, M., Precise tracking of the dynamics of multiple proteins in endocytic events, Methods in Cell Biology, Vol. 139, pages 51-68 (2017) if you use or modify this software.
You need Python 3.X. To install the spotquant package it is sufficient to download it here.
Open the downloaded folder in a terminal and type
sudo python setup.py install
The command python can be replaced by your Python 3.X installation of choice ( i.e. python3.5, python3.7, etc. etc.).
It is sufficient to write a .py script with the following two lines
import spotquant.measurespots as sq
sq.experiment( path = 'a string containing the path to the experiment folder' , target_name = 'Foo1' , reference_name = 'Foo2' , target_median_radius = 11 , reference_median_radius = 17 , save_masks = True , measure_max_intensity_frame = True )
The experiment folder must contain two subfolders.
One of these subfolders contains the images of cells expressing the target protein tagged with your fluorophore of choice, the other subfolder contains the images of cells expressing the reference protein tagged with the same fluorophore as the target proteins.
It is important that the folders are named with the target_name and reference_name that are entered as variables in the experiment function.
The script will output the fluorescence intensities as .txt files in the folder identified by path. The script creates a masks folder in each of the two subfolders that contain the raw images.
These masks folders contain all the masks used to identify and select the spots (*_PatchMask.tif and *_CtrlMask.tif), the images of the cells corrected for local background subtraction (*_BkgCorrected.tif); see Picco and Kaksonen, 2017 for details about the image background subtraction) as well as the median filtered image used to compute the background subtraction (*_Median.tif). These images are essential to asses the goodness of the parameters target_median_radius and reference_median_radius. A smaller median radius will leave the spots still visible on the *_Median.tif image. If that is the case, the radius must increase. Using the ImageJ preview, in Process > Filters > Median... is useful to find the ideal radius. A composite image of *_PatchMask.tif and its corresponding *_BkgCorrected.tif (in ImageJ: Color > Make Composite) will allow to identify the patches that have been selected:
(in red is the *_PatchMask.tif mask for that selected frame).
save_masks is an option to chose whether to save or not all the mask used to identify and measure spot intensities. It is set to True as default and creates a masks folder in the analised folder, which contains the images used for the analysis.
measure_max_intensity_frame defines how the fluorescence intensity of a spot is measured. If True, the fluorescence intensity is measured from the brightest frame in which each spot is imaged. That is the algorithm used in Joglekar et al., 2006 and Picco et al., 2015 and it is useful when the structures imaged are susceptible to move along the z-axis during the z-stack acquisition. With this option set as True, not all the photons captured by the objective are taken into account. When it is set to False, the fluorescence intensity is measured as the integral of the pixel values of each spot imaged in the different planes of the Z-stack.
spotquant has a basic function to estimate the number of molecules of the target protein spots, knowing how many molecules are present in the reference protein spots.
Load the fluorescence intensities measured in the patches, which are saved as a .txt file, with
import spotquant.measurespots as sq
foo_intensities = sq.load_quantification( "the_path_to_your_quantification/foo_intensities.txt" )
foo_reference_intensities = sq.load_quantification( "the_path_to_your_quantification/foo_reference_intensities.txt" )
or use directly the outputs of experiment, which outputs both the reference and target protein intensities (the same that are saved into the .txt file).
The number of proteins can be estimated as
foo_number_of_proteins = sp.quantify( foo_intensities , foo_reference_intensities , r_number )
where r_number is a tuple with two entries: the known number of proteins present in the reference patches, and its standard error.
Refer to example.py, which analyses the images in the example folder, for a test using both measure_max_intensity_frame = True (default) and measure_max_intensity_frame = False. The number of target proteins will be (257.65, 7.21) and (277.86, 10.04).
The fluorescence intensity values are computed as median (distributions are skewed). Their error is estimated with the MAD corrected for asymptotically normal consistency on the log transform of the raw fluorescence intensity values (used to aproximately conform to normality). The error associated with each fluorescence intensity value will thus be:
where are the measured fluorescence intensity values in each spot.
The error on the estimate of the number of molecules will be:
where is the known number of molecules in the refrence spots,
and
, are the median fluorescence intensity values measured for the reference and target protein respectively, and
,
are their respective error estimates derived from the log transformation described above.
is the error estimate on the known number of molecules in the reference spot.