Robust integration of single-cell cytometry datasets.
Technical variance is a challenge for working with large single-cell #cytometry datasets - which is why we are proud to present cyCombine as a tool enabling #DataIntegration both within and across technologies.https://t.co/G1Vs4KNWo5
— Christina Bligaard Pedersen (@cbligaard) March 31, 2022
cyCombine is composed of two main modules: One for #BatchCorrection and one for panel merging. In combination, they can yield a complete integration of data, also when the antibody panels are not identical.
— Christina Bligaard Pedersen (@cbligaard) March 31, 2022
In the article, cyCombine is applied to a dataset of CLL patients and healthy donors to demonstrate its full potential - and even with a large dataset of more than 12 million cells, cyCombine is able to run quickly on a standard laptop.
— Christina Bligaard Pedersen (@cbligaard) March 31, 2022
A comparison with other tools also underlines the flexibility and robustness of our new tool, which we are excited to share with the community as an R package on GitHub.
— Christina Bligaard Pedersen (@cbligaard) March 31, 2022
I am grateful to have worked with an excellent team of scientists on this project including @sorenhdam, @mikebarnkob, @JALALFA, @SatyenGohil, @larsronnolsen - and of course the non-Twitter users: Mike Leipold, Noelia Purroy, Laura Rassenti, Thomas Kipps, Jennifer Nguyen, Cathy Wu
— Christina Bligaard Pedersen (@cbligaard) March 31, 2022
Make sure to check out the R package and the vignettes.
Four years ago I had a beer with @larsronnolsen and complained about what a fuss it was to find good fluorochromes combinations for #flowcytometry. And today we wrote this: https://t.co/jlrb8layMC. It’s still a work in progress with functions being added, but here it is.
— Mike Bogetofte Barnkob (@mikebarnkob) March 19, 2021
The idea is quite simply that humans are pretty bad at choosing which fluorochromes to combine, especially when having to pick a large number: 5-10–15, not to say 20 or more. Which are best of these ten? pic.twitter.com/AgakFdh4xm
— Mike Bogetofte Barnkob (@mikebarnkob) March 19, 2021
So why not let a computer calculate the ones with the least overlap. Easy right? Well, it’s actually quite complex especially if you want to compare more then 30 different spectra. Fluorochromes these days are abound! We included over 100 on Spectracular, but there are many more
— Mike Bogetofte Barnkob (@mikebarnkob) March 19, 2021
Enter Alfredo, Gianfranco and Lars who came up with some brilliant ideas for how to turn this complexity into something that gives combinations that are near optimal suggestion, and does so in only a few seconds.
— Mike Bogetofte Barnkob (@mikebarnkob) March 19, 2021
We wanted to test Spectracular against something and decided to look at the expertly-made OMIP panels, which started out back in 2010. Turns out you can improve considerably on them if you start taking into account all the fluorochromes (=k in image) available today: pic.twitter.com/aO35ZEWdA7
— Mike Bogetofte Barnkob (@mikebarnkob) March 19, 2021
We are still working on my favorite function in all this: the ability to couple one or more antibody to several specific fluorochromes and ask Spectracular to find the best combinations of these and say five others. Stay tuned. But here’s the beta version: https://t.co/47xKSZdmw5
— Mike Bogetofte Barnkob (@mikebarnkob) March 19, 2021
Contact