Difference between "LogNormalize" normalization method in Seurat and z score normalization

Question

[f.morgana]

Hi all,

Perhaps a silly question but the default normalization method in Seurat is "LogNormalize". Instead of automatically running the code, I would like to understand what is happening when I perform the normalization and scaling steps in Seurat. And how exactly does this differ from z score normalization? How are the expression levels in e.g. the feature plots expressed?

Please take into consideration that I am not comfortable with statistics, so do not hesitate to try to give a simple and elaborate explanation!

Many thanks in advance, Florencia

Answer 1

[ATpoint]

LogNormalize scales the data by library size so it eliminates the confounding effect of different read depth across the cells and then applies log2 transformation.

Z-scoring (see Wikipedia article) is a metric that indicates how much a value deviates from the mean of many samples. In the (sc)RNA-seq context that would be how much the read count of one cell for a given gene deviates from the mean of the counts of all cells for that gene. Z-scoring is typically performed based on the log-normalized counts.