Take a look at HTX

[AnneCarpenter]

Please forgive me for a blatant mis-use of the Github issue system: this is not an actionable code issue; I wasn't sure how to reach those interested otherwise.

This software might be interesting to take a look at (in MATLAB though)...

bioRxiv paper: https://www.biorxiv.org/content/early/2017/10/16/204016.article-metrics Github: https://github.com/CarlosArteta/htx

[tischi]

Thanks for pointing this out to us! I am surprised to see MATLAB having a comeback :-)

In fact, there are several tools coming out right now:

• https://www.phaedra.io/
• TissueMap from FMI Basel (could however not find it on the www yet)

That's also why I mentioned the issue at the CZI meeting to see whether one can join forces at some point...but maybe this is too diverse of a field...anyway my plan for the next months is to get an overview of what's around and then take it from there.

For "downstream image-derived data analysis" I am however kind of favoring an option that uses R as its code base, because my experience is that if one uses another solution one has to in the end often combine it with R anyway. What's your take on this?

Maybe also interesting to work on a wish list for "the perfect tool"

• runs both locally and in www server mode
• different plotting modalities, e.g. scatter plots and heatmaps
• visual quality of plots should be publication ready
• linking numerical plots to images and segmentation overlays
• good scientific 5D image viewer (ideally showing raw data)
• offering different data normalization strategies
• easily expandable (i.e., if one specific statistical method does not exist it should be possible to add it within let's say 1 day without changing the code base)
• GUI for most commonly used functions
• CL API for expert users
• Command recording such that a particular analysis can be repeated, e.g. one script that can be run from Command Line, without the need to start the UI.

[imagejan]

@tischi wrote:

• TissueMap from FMI Basel (could however not find it on the www yet)

It's here: https://github.com/TissueMAPS (and it's more from Zurich, but being used at FMI Basel as well ;)

[AnneCarpenter]

Good point @tischi about R. Jupyter notebooks would be an awesome way to go.

[AnneCarpenter]

An additional wish list item:

• able to plot millions or even billions of cells (e.g., in a density plot)
• ability to export per-cell labels (e.g. if you make a gate in one of the plots, export which cells are inside vs outside the gate)

We could at some point re-read the CellProfiler Analyst papers and list what we described as must-have features in there but as far as I can recall, you have the major ones listed.

Note that Peter Horvath and I are talking about combining forces for supervised machine learning (our Classifier function of CellProfiler Analyst and his Advanced Cell Classifier tool). So if we are able to create a merged, better tool for the Classifier functionality, it leaves the exploration/visualization tools of CellProfiler Analyst orphaned and thus likely happily replaced by ShinyHTM or whatever you evolve into. We've not had much bandwidth to maintain CellProfiler Analyst so that could be a nice solution for all.