FAQ  •  Register  •  Login

Determining the divergence between samples

Forum rules
Please be as geeky as possible. Reference, reference, reference.
Also, please note that this is a mixed bag of math-gurus and mathematically challenged, so choose your words wisely :-)
<<

RSchackmann

Participant

Posts: 9

Joined: Fri May 27, 2016 5:26 pm

Post Fri Sep 09, 2016 12:17 am

Determining the divergence between samples

I've been searching for a way to give a score to the divergence between two samples. In the Amir & Pe'er Nat biotech 2013 paper (viSNE PMID: 23685480), they use the Jensen-Shannon (JS) divergence to compare the viSNE divergence of two/three healthy samples and state a JS-divergence of 0.04.
I've been trying to reproduce this in Dana Pe'er's analysis software Cyt(Sightof). I made viSNE maps of two very different samples (sample A and sample B)
AvsB.jpg


and use the 'compare maps' option in cyt as indicated in the Cytutorial.ppt.
compare_maps_V2.jpg
compare_maps_V2.jpg (63.05 KiB) Viewed 7700 times

In the window that pops up I can however only select the viSNE maps of one file, and not compare two files to each other.
Hence, I've tried to compare the viSNE map of sample A to the viSNE map generated from both sample A and sample B, but the divergence score is only ~0.1 even though the viSNE maps show no overlap between sample A and B; so that is clearly not the way to do it.
Options_A_V2.jpg

The paper 'supplementary figure 5 A' shows 3 marrow donors; clearly one viSNE was generated on all three marrows and then the JS divergence was calculated (where they conclude the samples are highly similar with a divergence of ~0.04). I thought they used Cyt, but I'm starting to doubt that now..

My question is, does anyone know how to determine the similarity between two samples through viSNE or other means (without knowing much about R...)?


Thanks for your time,

Ron
<<

margotvanspronsen

Participant

Posts: 1

Joined: Tue Sep 27, 2016 10:56 am

Post Tue Sep 27, 2016 5:47 pm

Re: Determining the divergence between samples

Hi,

Unfortunately I've no answer, but exactly the same problem. I was wondering whether you already known how to apply the JSD using Cyt tool or another software program in order to compare visne maps. Hope to hear from you!

Margot
<<

RSchackmann

Participant

Posts: 9

Joined: Fri May 27, 2016 5:26 pm

Post Wed Sep 28, 2016 6:22 pm

Re: Determining the divergence between samples

Hoi Margot,
I contacted the cyt team, and they told me it is not possible to compare maps between samples in cyt. I'm in contact with the first author on the visne paper, and am right now looking into a way to create a probability distribution of a 2d map (binning the sne1 and sne2 'coordinates' into a single value), but haven't figured it out quite yet...

I'll let you know once I know more.

,
ron
<<

dtelad11

Master

Posts: 129

Joined: Mon Oct 31, 2016 6:26 pm

Post Wed Nov 02, 2016 3:33 pm

Re: Determining the divergence between samples

Following an email correspondence with RSchackmann I implemented the analysis from the 2013 viSNE paper as an R script. Unfortunately, I do not know how to post files to the forum. We are currently working on setting up a github repository for the Human Immune Monitoring Core (HIMC) at Mount Sinai. Until then, you're welcome to contact me at el-addavid.amir@mssm.edu for the script.
<<

RSchackmann

Participant

Posts: 9

Joined: Fri May 27, 2016 5:26 pm

Post Wed Nov 02, 2016 6:45 pm

Re: Determining the divergence between samples

You beat me to it El-ad, I was just writing a post.

I could upload the script you made, but it might be wiser to put it on github and keep it centralized;

Let me know if you want me to upload it here!
<<

dtelad11

Master

Posts: 129

Joined: Mon Oct 31, 2016 6:26 pm

Post Wed Nov 02, 2016 7:06 pm

Re: Determining the divergence between samples

Thanks RSchackmann!! We're in the process of setting up a public github repo for the HIMC, I'd much prefer to have the script available there.
<<

sfauteux

Participant

Posts: 17

Joined: Tue Feb 23, 2016 2:18 pm

Post Mon Nov 07, 2016 2:55 pm

Re: Determining the divergence between samples

It is possible to calculate Jensen-Shannon divergence in Cyt but it requires a little bit of annoying tweaking.

Subsample your files at a given # of cells.

Run a comparative Visne on your subsamples with the paramaters of interest.

Save this new FCS file. This will contain the x and y parameters of SNE. Convert in CSV, copy paste in a new CSV the column of xSNE and ySNE (give recognizable headers). You will have 2 column per file.

Convert CSV in FCS

Import the new FCS in cyt.

Compare your maps with eachother.

Compile.

S

Return to CyTOF data analysis

Who is online

Users browsing this forum: No registered users and 10 guests