FAQ¶
What preprocess should be done on the scRNA-seq data?¶
Cyclum is noncommittal on what specific transform is used, as long as the circular pattern is preserved. That said, experiments show that scaled (variance = 1), centered (mean = 1) log-transformed (\(\log_2(x+1)\)) data work well. Using Freeman-Tukey transform (\(\sqrt{x} + \sqrt{x+1}\)) may also be a good idea.
How do I choose number of linear dimensions?¶
We find 0 or 1 generally works well when cell cycle is really a problem in the data.
What does the pseudo-time mean?¶
It is the inferred stage of a cell. One pseudo-time for one cell itself does not make sense because it is not a real time. However, pseudo-times for more cells reveals their order and similarities. The pseudo-time for many cells shows the trajectory of cells in a scale.
What does the circular pseudo-time mean?¶
It means we assume that there is no start or end in the time. Cells are going through a circular process, presumably cell cycle. That said, prior knowledge can be used to determine which pseudo-time is G1 phase and M phase, and they can be considered the start and end.