Does Seurat have Poisson normalization method?

Question

[Emily]

Does the current Seurat version have Pearson residual normalization? If not, is there an alternative tool to do scRNA analysis with Pearson residual normalization?

Answer 1

[antonioggsousa]

Hi,

Yes, it has the sctransform method:

• paper: Hafemeister and Satija, 2019

• vignette: sctransform pipeline

The Hafemeister and Satija, 2019 clearly states in the abstract (citing):

"We propose that the Pearson residuals from “regularized negative binomial regression,” where cellular sequencing depth is utilized as a covariate in a generalized linear model, successfully remove the influence of technical characteristics from downstream analyses while preserving biological heterogeneity."

I hope this helps,

António

P.S.: Just came across the following paper - Ahlmann-Eltze and Huber, 2022 - comparing several transformations for single-cell RNA-seq. Under the scenarios tested the simplistic log1p normalization method with a scaling factor (i.e., log(y/s + y0)) outperforms more sophisticated methods taking into account the metrics assessed.

[Emily]

Thank you. Follow up question Seurat doesn't use Poisson method does it? I was looking at this paper and was trying to analyze my data using Poisson method https://genomebiology.biomedcentral.com/articles/10.1186/s13059-021-02451-7

[antonioggsousa]

Sorry, I don't know the method to answer your question.

What it is clear from the description above is that the Pearson residuals come from a negative binomial regression.

[Emily]

Understood. Thank you for your help!