hello again,

i was wondering whether you guys are also planning to support some “unsupervised learning” in the new version of CPA.

i guess the “Gating” that is present in the current version of the software is some kind of clustering, but only based on a subset of two features.
i could image, it may sometimes be useful to construct a more involved “metric” that measures how close or distant two images are in the feature space…and then show the result as some sort of tree-diagramme. although, there may be other software to do such kind of things…?!

any thoughts on this?

best wishes,

We haven’t explicitly spoken about doing unsupervised classification in the next version of CPA. Most of the features we implement come out of satisfying some specific need of our collaborators, and apparently clustering etc hasn’t come up yet (I think because we’ve been able to adapt the classifier to many different situations). At any rate, I’ve done some research in the field of competing analysis software, and there is only one other app out there for that I know of that is geared towards biological image based screens. It’s called AcuityXpress, by Molecular Devices ( … press.html scroll down to Multivariate Analysis). They actually implement a handful of different clustering methods.

I am currently the only person focusing efforts on the development of the next generation of CPA. As I stated before, things are looking very good for the next gen classifier, and we do anticipate releasing that separately from the rest of the software since most of our collaborators have used CPA solely for the classifier. As development proceeds, we will be considering how to implement visualizations useful for image based HTS. We’ll likely leave room for data clustering, but as of now, we’re mostly interested in allowing users to do grouping more categorically (as could be translated to a SQL select statement).

I just would like to know if you consider developing unsupervised machine learning modules for CPA at this moment. I’m sure this would be very helpful sometimes for scientist non-well-known with programming seeking for some not easily defined by eye cell types in co-culture experiments.
Thank you very much for your software and looking forward for upgrades.

Hi Vitalii,
Unfortunately, CPA is no longer undergoing active development at this time; see here: Important CellProfiler Analyst Announcement. If some enterprising individual is willing to contribute code, we’re more than happy to review it!