Single Cell RNA seq imputation with MAGIC - scale of output?

Question

[dps2501]

Hello,

I ran MAGIC on some single cell seq data and am curious of the scale of the output? I first used Seurat's normalization function which converts the raw counts to CPM and then natural log transforms the data. After applying MAGIC is the data on a linear scale or natural log scale?

Thanks

Answer 1

[Friederike]

What does it look like when you plot the distribution of values?

Based on the comments of the author here, I assume that the scale of the values should not have changed (there should just be a lot fewer zeros), BUT there may be small negative values because "after MAGIC the imputed data (imputed PCs) are projected back to the original space and as a result there can be (very small) negative values" github issue.

[dps2501]

Thank you! I should have searched a little more on my own. I looked at the distributions before and after imputation and they appear to be on the same scale.

[Friederike]

No worries, I remembered that I had the same question way back and therefore knew where to look for the answer