histoCAT++ manual

Looking for ways to analyze your dataset? Need help with a software package?
Computational image analysis of multiplexed Imaging Mass Cytometry acquisitions
pbull
Posts: 4
Joined: Mon Jul 16, 2018 6:40 pm

Re: histoCAT++ manual

Post by pbull » Mon Jul 16, 2018 6:44 pm

What are the computer requirements needed to efficiently create t-SNE plots? Our normal desktops are incapable of handling even the sample data set given on the histoCAT website.
ssivajothi
Posts: 33
Joined: Thu Jan 11, 2018 9:00 pm

Re: histoCAT++ manual

Post by ssivajothi » Tue Jul 17, 2018 7:02 pm

Hi Raul,

I have a question about exporting images from histocat++. I imposrt the .txt file, apply compensation and want to export. I can see visual expression of markers for some channels. When I perform the export (stack for neighbors), the resulting tiffs all look black unless I go into image editing software and play with the levels. Do you know what is causing this? Is there some setting to adjust? And if I adjust settings like max, offset, mult etc on histocat++ is this carried forward in the export? Please let me know.

Santhosh
RaulCatena
Posts: 30
Joined: Mon Nov 27, 2017 12:49 pm

Re: histoCAT++ manual

Post by RaulCatena » Tue Jul 17, 2018 9:25 pm

Hi pBull,

The limiting factor is the processor's speed. The standard tSNE implementation uses only one core, so pretty much clock speed (as long as the dataset fits in memory which most likely will, specially if you don't have too many other apps open and the system does not need to keep swapping memory). So clock speed is the main factor. There is a Matlab/CUDA implementation at van der Maaten's website that you could put in the code to substitute the current implementation and if using a computer with Nvidia GPU would make it at least an order of magnitude faster.

Also, keep in mind that tSNE does not scale well. Double the cells and the time increases a much greater fold. In the order of a few thousands (1000 to 5000) is rather fast. In my laptop, 1 to 10 minutes. When I run 50000 cells it takes 14 hours.

Regards
RaulCatena
Posts: 30
Joined: Mon Nov 27, 2017 12:49 pm

Re: histoCAT++ manual

Post by RaulCatena » Tue Jul 17, 2018 9:31 pm

Santosh,

The reason is that data is saved as 16 bit tiff. So if you have channels that go to a few counts, it will look pretty much black if no automatic adjustment is applied. If you open the files with preview they'll look black. If you open them with Fiji, for instance, and let it use automatic adjustment, then it will fit the histogram range for that channel in a 8-bit visualizable scale, and you will be able to see the signal. histoCAT++ also does automatic adjustments and that's why you see the data clipped to its max right away. For subsequent processes like segmentation in Ilastik/Cell Profiler etc, this has no effect whatsoever, since the values in the file are simple the intensity number values and all decent image software packages can deal with that.

Hope this answer your question.

Regards,
ssivajothi
Posts: 33
Joined: Thu Jan 11, 2018 9:00 pm

Re: histoCAT++ manual

Post by ssivajothi » Wed Jul 18, 2018 3:06 pm

Hi Raul,

Thanks for that info, I did have a suspicion that it was the default processing by the software which was making it look black.

However, when I open .txt files in H++, some channels look completely dark, but when I bring the 'Mult' slider way down, I see the expected signal. Does this simply mean that my titration is low and I need to increase the amount of antibody? Will this type of channel work for analysis (since I assume the signal is there, just low) or better to add more antibody so that I can see signal without any adjustments?

Your input always appreciated.

Thanks,
Santhosh
ssivajothi
Posts: 33
Joined: Thu Jan 11, 2018 9:00 pm

Re: histoCAT++ manual

Post by ssivajothi » Mon Oct 29, 2018 12:50 pm

Hi Raul,

Just wanted to ask if there are any updates to Histocat++ and the associated manual?

Thank you,
Santhosh
Post Reply