Moving from proprietary software to CellProfiler for HCS


My facility uses the GE IN Cell 2000 High Content Screening microscope and we use GE software (IN Cell Investigator) for our subsequent analyses. I was wondering whether anybody else has successfully moved away from commercial software for High Content Analysis and if so, what their experiences were. Our screens are sizeable and extremely varied and the one thing the commercial package has going for it is that support is readily available and the relative ease of use. My background is biology, so hardcore programming is definitely not my strong suit. I’m just not sure whether CellProfiler will do the job and whether I’ll be able to get the hang of it well enough. Thanks.



Hi Eric.

We’re in a very similar position. The scientists use a number of packages to analyse their images and we’re seeing that the comercial packages don’t really scale. This is one of the major drivers for us to use CellProfiler. The concept of a scientist developing a pipeline on their laptop which can then be used to process 100s of 1536 well plates on a cluster is very attractive. Some colleagues have already run some CP screens of this size. One thing that you will need to do is put some infrastructure in place to make this accessible to the scientists as the vast majority of people working in HCS have zero Linux experience.

Most of the third party software stores data from image analyse in their databases etc. If you move to CP then you’ll need to find a way of keeping track of the data or you’ll risk ending up in “Excel Hell”. CP supports writing to databases, we’ve found that trying to save results in the middle of a run on the cluster isn’t reliable (too many processes trying to insert data at once, not CP’s fault), so getting CP to write the SQL files out and then doing a more ordered write is much better,

One thing that you may find is that some of the less technical scientists find moving away from their existing software a challenge. There a few things that they miss from the InCell software. One thing is the terminology ie nucleus, cytoplasm is more “biologist friendly” than terms such as primary objects. The other thing is that CP doesn’t currently have a way of viewing all of the images from a plate at once and it’s hard to select a few images from a plate to look at (ie +ve & -ve controls). I know the Broad are in the process of improving this area so it might become a non-issue soon.

We’ve also had very mixed experiences with support from the various vendors. I’ve found that I can usually get answers from this forum, I think our scientists/IT infrastructure team use the vendors’ support for troubleshooting badly-behaved application servers and databases rather than for “help me segment this cell” type questions.

In terms of the results from CP Vs the commercial tools, we’ve found that in some areas CP performs the best, in others one of the other packages shines. We seems to be able almost everything we want to do in CP.

If you’ve got any questions then please let me know.