This code is associated with the paper from Bhatia et al., "Quantitative analysis of auxin sensing in leaf primordia argues against proposed role in regulating leaf dorsoventrality". eLife, 2018. http://dx.doi.org/10.7554/eLife.39298

Authors and contact:

Neha Bhatia (bhatia@mpipz.mpg.de), Henrik Åhl (henrik.aahl@slcu.cam.ac.uk), Henrik Jönsson (henrik.jonsson@slcu.cam.ac.uk), Marcus Heisler (marcus.heisler@sydney.edu.au)

The analysis and figure creation can be replicated by running the scripts

code/deconvolution.py
code/segmentation.py
code/figure_generation.py

in succession, after required packages are installed (see below), and the 'project_path' variable has been updated (at the top of each script) to fit with your installation. Note that the deconvolution uses large amounts of RAM, which might prove troublesome for small-scale desktop computers.

For simplicity, all original data, intermediate data, and figure files used for quantifications are also provided within this repository and files are organised as follows:

code: holds the python source code, where the main scripts are deconvolution.py,
segmentation.py, and figure_generation.py. The environment.yml file can be used if
installation is done via Anaconda (see below).

code/external: local version of external packages needed to run the scripts (see below).

figures: all figures produced by the scripts (figure_generation.py) in pdf and 
png formats

intermediate_data: Holds the (manually cropped) stacks for individual leaves
before and after applying the deconvolution.py script. The files produced by the deconvolution 
are input files to the segmentation.py script.

misc/laser_settings.txt: a text file with confocal settings used by the deconvolution.py 
script

processed_data: holds the extracted data from the segmentation.py script, and also
stacks of tiffs showing the segmentation result where each identified nuclei has a specific
label for its voxels.

raw_data: contains the original data in lif and tiff formats. Note, all original data 
for the submission is provided via https://www.ebi.ac.uk/biostudies/studies/S-BSST223.

For queries relating to the paper, contact Marcus Heisler (marcus.heisler@sydney.edu.au). Questions related to the code are best addressed to Henrik Åhl (henrik.aahl@slcu.cam.ac.uk).

.tar.bz2 files must be decompressed before scripts can be run on the contained data:

$ tar -xjf raw_data.tar.bz2
$ tar -xjf intermediate_data.tar.bz2

We utilized Scikit-Image (0.14.1) for several of the image quantification steps. However, the distribution as-is contains a bug which causes problems when deconvolving images which contain values close or equal to 0. We therefore corrected this part of the code by using the machine epsilon for close-to-zero values. Specifically, the change consists of exchanging the code snippet

relative_blur = image / convolve_method(im_deconv, psf, 'same')

which is found on on line 389 in skimage/restoration/deconvolution.py, with

eps = np.finfo(image.dtype).eps
x = convolve_method(im_deconv, psf, 'same')
np.place(x, x==0, eps)
relative_blur = image / x + eps

A modified version ready for install is provided in code/external/scikit-image-0.14.1. For installation:

cd code/external/scikit-image-0.14.1
python setup.py install
cd -

PyCostanza is used for segmentation. For installation of the locally provided version:

cd code/external/pycostanza-0.1.3
python setup.py install
cd -

PyCostanza can also be installed via pip by running

pip install pycostanza==0.1.3

For installation of PSF using the locally provided version:

cd code/external/psf-2018.02.07
python setup.py build_ext --inplace
python setup.py install
cd -

In order to run the code and replicate the output, install required packages by running:

pip install numpy pandas scipy tifffile mahotas==1.4.5 matplotlib==2.2.3 scikit-learn==0.20.0

If you are running Anaconda, you can find an environment file in code/environment.yml. Install the environment by running

conda env create -f code/environment.yml

from the repository root directory. The custom packages will have to be installed as described above.