Seurat "Expression Level" units

Question

[paulranum11]

I am trying to understand what the correct units are for the violin plots output by Seurat. The Y axis is labeled "Expression Level" by default on their violin plots. If I input a matrix of counts values will my units then be log counts? likewise, if i input a matrix of TPM values will the units be log TPM?

Or alternatively are the units changed by the internal Seurat normalization process? Seurat (v1.4.0.8) has normalization process run using setup.

Setup(object, project, min.cells = 3, min.genes = 1000, is.expr = 0, do.logNormalize = T, total.expr = 10000, do.scale = TRUE, do.center = TRUE, names.field = 1, names.delim = "_", meta.data = NULL, save.raw = TRUE)

Some of the arguments for setup seem to include additional data normalization steps.

do.logNormalize = whether to normalize the expression data per cell and transform to log space.

do.scale = In [e-mail], perform row-scaling (gene-based z-score)

do.center = In [e-mail], perform row-centering (gene-based centering)

What are the appropriate units for data imported in this way?

NOTE: yes i know Seurat (v1.4.0.8) is outdated. But i am trying to understand data that was processed using this version.

Answer 1

[Friederike]

I'm assuming that the behavior did not change in Seurat v2 -- in Seurat v2, the data stored in the data slot (not the counts, which are typically stored in raw.data) are used for the visualizations, and that slot will only be filled if you used the normalization parameters you mentioned above. Santosh, another biostars user, pointed me to this helpful FAQ page that explains the three different types of data that Seurat stores if you follow their standard workflow (Answer 7). The slot names may have changed in v2, but conceptually, I don't think Seurat would have ever used raw counts for the types of visualization that you mentioned.