Why different tSNE plot with Seurat and Monocle if they'r using the same R package?

Question

[jgarces]

Hi all,

I'm doing my first serious analysis with single-cell data and I've a (maybe naïve, sorry) question.

I've analysed the data with Seurat and Monocle R packages... when I draw the tSNE plot for my sample, they are totally different and I don't understand it because internally both are using Rtsne package with defaults parameters. Any idea of this differences?

[Lluís R.]

Each time you run tSNE you'll get different results. I think you should be able to see this if you do another plot with Seurat and Monocle, you should see that it would change (unless they keep the same seed and random numbers). Also are you sure it is the same data? The first one is painted with two colors and the second one with 3.

Answer 1

[dppb05]

Despite both Seurat and monocle using Rtsne there are a few reasons the plots you got are different:

• Assuming you have used their respective standard pipelines on your data: they have different pipelines which will alter the data considerably, especially in the QC part.

• They are using different seeds: in Seurat::RunTSNE you can change the seed from the default value of 1, whereas in monocle::reduceDimension the seed used is 2016. You could try changing Seurat's seed to 2016 and see if you can get something closer to that of monocle, although, due to the previous point, it may not happen.

Lastly, keep in mind that in Seurat::RunTSNE you can change the parameter tsne.method in order to use a different function other than the default Rtsne.

[jgarces]

Thanks both, guys! Nothing simpler that set.seed()...