Question: What is t-Distributed Stochastic Neighbor Embedding (t-SNE)?
Answer: t-SNE is a probabilistic method for visualizing high dimensional data. t-SNE in our Single Cell analysis entails gene expression measurements per cell in a low n-dimensional space (n=2 by default in cellranger count
).
The n-dimensional plot is constructed such that if two points/cells are close, they are most likely to be close in the higher dimensional space as well, while if two points are farther apart they are likely to also be farther apart in the original higher dimensional space. Only the relative distances between points matter - the two axes are arbitrary and do not have any specific interpretation.
There are a number of t-SNE tutorials and references on the web, such as: