Question

analysis of FACS data using BioC

1

Entering edit mode

Bogdan ▴ 670

@bogdan-2367

Last seen 6 months ago

Palo Alto, CA, USA

Dear all,

we'd appreciate to have your advice please on the FACS analysis packages :

we do have the results of FACS experiments after PI staining in order to map the cell-cycle phases (in basal state, and after drug inhibition), and we'd like to ask you please:

<> which BioC packages for FACS data analysis would you recommend please ?

<> referring to the data presentation, shall we present it as COUNT HISTOGRAMS, RELATIVE HISTOGRAMS, DENSITY PLOTS, eCDF ?(as in some experiments, we do have a different number of cells that were PI-stained).

thanks a lot,

-- bogdan

facs prada flowstats • 1.6k views

ADD COMMENT • link 3.6 years ago Bogdan ▴ 670

score 1 · Answer 1 · 2020-09-29

1

Entering edit mode

helucro ▴ 70

@crowellh-11823

Last seen 8 months ago

University of Zurich, Switzerland

Dear Bogdan, I don't know the details of your biological question, but would recommend

flowCore for basic data import and transformation
flowWorkspace and openCyto for gating (e.g., on FSC, SSC, or to identify peaks of PI) and some simple readouts (e.g., fraction/number of selected cells);
and, most definitely, ggcyto for neat visualizations (e.g., histograms, 1/2d densities, split by sample/condition) with options to include gates and annotations of e.g. cell counts/proportions

Again, while I am not 100% sure what exactly you were looking for, I would say these are probably the most widely used Bioc packages in this context and a great place to start!

ADD COMMENT • link 3.6 years ago helucro ▴ 70

1

Entering edit mode

Thanks for mentioning our cytometry packages. We are always making improvements and trying to decrease the learning curve for new users. To that end, it's still in its infancy, but you can check out cytoverse.org (the cytoverse is the collective name for all of those tightly-coupled cytometry analysis packages. The vignettes for each package are a good place to start, but we will also be adding examples there of how to perform common and not-so-common tasks.

I'd also recommend checking out CytoExploreR, which uses those packages as a core but adds some additional stuff, including some nice GUI elements.

ADD REPLY • link 3.6 years ago Jake Wagner ▴ 310

0

Entering edit mode

Thank you Jake. If I may ask for your opinion too, on the question that 've posted a few minutes ago above your message please. I will explore in more detail : https://dillonhammill.github.io/CytoExploreR/

ADD REPLY • link 3.6 years ago Bogdan ▴ 670

1

Entering edit mode

I agree with Helena's points. Particularly, if you plan on pooling data from multiple samples, that downsampling to equal representation could be a good idea. If you'll be looking at separate plots for each sample, though, that's not really an issue.

Regarding density plots vs frequency histograms, it's sort of up to you. It also depends a little bit on how abundant the data is/how well it covers the range of the feature, as that can sort of effect how well the KDE-smoothing of the density plots captures the information in the histogram. Basically, a few sparse events in small bins can cause little bumps in the density plot that are harder to visually quantify than a simple histogram bin.

ADD REPLY • link 3.6 years ago Jake Wagner ▴ 310

0

Entering edit mode

thank you Jake ! if I may add please, talking about the number of cells : nowadays, how many cells would you recommend to be collected per each experiment ? (we will down-sample, although doing a repeat experiments also helps to confirm).

ADD REPLY • link 3.6 years ago Bogdan ▴ 670

0

Entering edit mode

Thanks a lot, that is very helpful ! If I may add a question please : in some experiments, my colleagues have collected different number of cells (for example 30 000 cells in an experiment, and 100 000 cells in other experiments). In this case, shall we rather show the data on DENSITY PLOTS rather than on COUNT HISTOGRAMS (for a specific marker) ? thanks again !

ADD REPLY • link 3.6 years ago Bogdan ▴ 670

1

Entering edit mode

Yes, absolutely!

Generally, if you believe large differences in the number of cells will have any downstream effects you could downsample samples and/or conditions to have comparable sizes. But for cell cycle analysis this should not be the case - I think.

For any plots NOT separated by sample/experiment, however, I would be cautious. Say you visualise all cells together in a 2D density plot / scatter. Then, the significantly larger one will "take over" the plot, i.e., you will see a high density for one sample/experiment, while the other will have a density of virtually 0 in comparison and become invisible. Maybe something to keep in mind...

ADD REPLY • link 3.6 years ago helucro ▴ 70

0

Entering edit mode

thanks a lot, Helena ! very helpful indeed ! just to double check and confirm with you please : shall we have FACS data of PI staining with different number of cells in two experiments (eg 30 000 and 100 000, respectively), using the DENSITY PLOTS would be well accepted data (rather than FREQUENCY HISTOGRAMS) ? (my colleagues did the FACS experiments ; myself, I did the last FACS experiments in 2008 :)