FAQ  •  Register  •  Login

Papers that exclude markers from viSNE analysis




Posts: 3

Joined: Thu Sep 17, 2015 2:17 pm

Location: Bergen, Norway

Post Wed Feb 10, 2021 2:43 pm

Papers that exclude markers from viSNE analysis

Hi all,

I'm trying to figure out how to respond to a reviewer. The question is about why I stained with 36 markers but am only using 26 markers for the viSNE analysis. I've been doing CyTOF for a long time and to me, this is very normal and there are various reasons to do this. I have explained that markers were excluded because they were used for pre-gating, are negative or very low, or are activation markers and the point of this viSNE was phenotyping (activation state to be assessed later on).

I'd really like to be able to point to CyTOF publications (other than my own) that make note of leaving markers out of the analysis or that specifically discuss the practice. Diggins, Ferrell et al. 2015 discusses leaving out negative markers from the analysis but doesn't show any data. Many papers don't specify which markers were used in viSNE, SPADE, CITRUS, etc. The original viSNE paper (El-ad et al. 2013) excludes markers but only to show that viSNE still works and not because the viSNE map will be better without these markers.

Any suggestions for papers that might help would be much appreciated!




Posts: 94

Joined: Wed Dec 21, 2016 9:22 pm

Location: Marseille, France

Post Wed Feb 10, 2021 3:35 pm

Re: Papers that exclude markers from viSNE analysis


The usual goal of a tSNE is to show populations, which is achieved by the use of the lineage/type markers. The split between type and state markers has been discussed in the F1000 article/pipeline of Mark Robinson's lab or in its reviews. This subject is quite sensitive for some people.

IMHO it does not make sense to merge all markers in a single tSNE unless your expertise says it does.
a) in the core of any current numerical approach there is a distance calculation that uses all the given markers at once. Let's consider 2 cells. If there are many markers that are different, great, the two cells will be identified as different. What if there is only 1 marker pointing a difference while the 25 others are just bringing noise? Will we hear its voice in the brouhaha? This is a question of curse of dimensionality, as discussed by Newell et al. IIRC.
b) if it was so obvious and useful to merge all markers, I don't think E. Newell would have published OneSense.

Your explanations are crystal clear and obvious. I am wondering why a reviewer asks such a question. Let's imagine it is curiosity. Send your answers, bring the tSNE with all markers, but don't add it to your article if you don't feel it should.


Return to CyTOF general discussion

Who is online

Users browsing this forum: AdsBot [Google], Bing [Bot] and 1 guest