May 25, 2018  |  Education, Events, viSNE, Webinars  |  By  |  0 Comments

Webinar Recap: Advanced viSNE (t-SNE) Review

Untitled-1Nearly 500 scientists registered for our latest data analysis webinar. Cytobank Application Scientist Geoff Kraker took us on a tour of recently added features like setting the seed (for replicating results), configuring display gate labels, and more. If you missed it, not to worry. The recording is now available online.

cyobankarrow40x40View our latest webinars on the Cytobank YouTube channel

Following the demo, Geoff fielded many great live questions from the viewers. Here are some of our favorites:

Q: How do you run a SPADE or CITRUS on your viSNE?

A: After you’ve completed a viSNE run thSpadeOnviSNEat you’re satisfied with, from within that viSNE experiment start a new SPADE or CITRUS analysis. Then you can set up SPADE or CITRUS as normal, typically using the same clustering channels as you used for viSNE. The t-SNE1 and t-SNE2 channels will be in your data now so that you can use them for color overlay dot plots to show your results. Alternatively, instead of selecting a whole series of surface markers, you could choose to select only t-SNE 1 and t-SNE 2, as shown in the example on the right. This will basically split apart your viSNE map by clusters. Learn More

Q: What is the difference between viSNE and SPADE?

A: viSNE is a dimensionality reduction algorithm and SPADE is a clustering algorithm, which means that SPADE will put events into nodes or clusters, whereas viSNE rearranges the events on a map so that the similar ones are close together. Because of SPADE’s clustering, you can calculate statistics on the clusters output by SPADE or do fold change calculations between files as part of its results. In contrast, comparing statistics or files from a viSNE is a multi-step process, but it’s a great visualization tool. We typically run viSNE and SPADE (or CITRUS) in a pipeline, like the one discussed in our recent CITRUS webinar. Learn More

Q: If I run batch controls like the ones you mentioned and then realize that I do have batch effects, is there a good way to control for them?

A: Depending on the setup of the experiment and the control samples available, we at Cytobank typically run our own batch normalization to deal with batch effects. At this point, this normalization process is something that our team can provide as a part of our consulting services projects for customers. We would be happy to chat about this if you are interested. Learn More