Quantification of membrane and cytoplasmic concentrations based on differentiable simulation of cell cortex images
Project description
DISCCo: Differentiable Image Simulation of the Cell Cortex
Quantification of membrane and cytoplasmic concentrations based on differentiable simulation of cell cortex images. Designed for use on images of PAR proteins in C. elegans zygotes.
This extends on the segmentation and straightening algorithm described here, and uses similar underlying methods.
Methods
Our method is adapted from previous methods that model cross-cortex intensity profiles at each position around the cortex as the sum of distinct cytoplasmic and membrane signal components (Gross et al., 2018; Reich et al., 2019). Typically, these two components are modelled as an error function and Gaussian function respectively, representing the expected shape of a step and a point convolved by a Gaussian point spread function (PSF) in one dimension. In our model we relax these assumptions to account for the possibility of a non-Gaussian PSF and complex light-scattering properties which cannot be captured with these simplistic descriptions. Instead, cytoplasmic and membrane signal profiles are modelled as arbitrary vectors of length 50 pixels which can take on any shape (smem and scyt). Full straightened images can then be simulated as the addition of two tensor products:
sim = ccyt ⊗ scyt + cmem ⊗ smem
where ccyt and cmem are cytoplasmic and membrane concentration profiles. Building the model using the differentiable programming language JAX allows input parameters to be iteratively adjusted by backpropagation to minimise the mean squared error between simulated images and ground truth images:
In doing so, both the image-specific concentration parameters and the underlying quantification model (i.e. smem and scyt) can be optimised, allowing for closer simulations and more accurate quantification:
For full details of the method, see
PAPER IN PREP
Installation
pip install discco
Instructions
Binder link (TO DO)
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.