Histocat - Differential cell neighborhood analysis

Looking for ways to analyze your dataset? Need help with a software package?
Computational image analysis of multiplexed Imaging Mass Cytometry acquisitions
Post Reply
User avatar
sbooth
Posts: 8
Joined: Wed Apr 25, 2018 9:40 pm

Histocat - Differential cell neighborhood analysis

Post by sbooth » Tue Apr 23, 2019 11:36 pm

Hi all, hope you are well!

We have 70 samples analyzed by IMC, and they are graded 1-4 in disease severity. Is there a way to see if cell types interact differently between the severity stages?

For example, should we tsne+phenograph on all the samples together, and then run the neighborhood analysis on the different severity groups separately? So we would have 4 neighborhood analyses, then compare the significant interactions among them somehow.

Our hypothesis is that inflammatory cells may interact with structural cells more as the disease severity increases, leading to tissue destruction.

Any thoughts on this would be much appreciated!
Best,
Steve
DenisSchapiro
Posts: 36
Joined: Wed Nov 29, 2017 11:34 am

Re: Histocat - Differential cell neighborhood analysis

Post by DenisSchapiro » Wed Apr 24, 2019 2:57 am

Hi Steve,

I would run tSNE + PhenoGraph on all samples and run neighborhood analysis (1) on 4 different disease states for comparison and on all disease states together. When you run all together, do you see hierarchical clusters enriched in certain cell-cell interactions? When you run them individual, you can compare the heat maps. We have done something similar in the original histoCAT paper.

Best

Denis
User avatar
sbooth
Posts: 8
Joined: Wed Apr 25, 2018 9:40 pm

Re: Histocat - Differential cell neighborhood analysis

Post by sbooth » Tue Jul 23, 2019 1:11 am

Thanks Dennis,

I see in the saved neighborhood matlab data file there are a few matrices of interaction data. Am I correct that I would want to use "Matrix_Delta_cut_noNaN" to reproduce the clustergram?

Also, how do you retrieve the specific number of interactions between two cell clusters. i.e. for each sample how many interactions between cell phenotypes 1_2??

Thanks again,
Steve
DenisSchapiro
Posts: 36
Joined: Wed Nov 29, 2017 11:34 am

Re: Histocat - Differential cell neighborhood analysis

Post by DenisSchapiro » Wed Jul 31, 2019 7:38 pm

Dear Steve,

This is a pretty unique and tailored question. I will respond via email with your data example.

For the Clustergram, use the command: view(Delta)

Best

Denis
Post Reply