# Using pearson residuals for differential gene analysis

**URL:** https://discourse.scverse.org/t/using-pearson-residuals-for-differential-gene-analysis/1485
**Category:** scanpy
**Created:** [June 5, 2023, 8:35am UTC](https://discourse.scverse.org/t/using-pearson-residuals-for-differential-gene-analysis/1485 "2023-06-05T08:35:59Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![jeji0209](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/jeji0209/32/685_2.png) [@jeji0209](https://discourse.scverse.org/u/jeji0209)
#### Post date: [June 5, 2023, 8:35am UTC](https://discourse.scverse.org/t/using-pearson-residuals-for-differential-gene-analysis/1485/1 "2023-06-05T08:35:59Z")

</div>

Hello. I have 6samples (3 conditions, 2 replicates for each condition).  
I am trying to conduct differential gene analysis between the conditions.  
One of my sample has a significantly low count overall, so I would like to remove this variation, utilizing `sc.experimental.pp.normalize_pearson_residuals()`.  
However, I am not confident if it is okay to use pearson normalized matrix for DGE analysis.  
I notice that this normalization method is similar to Seurat’s SCTransform, but SCTransform datas go through `PrepSCTFindMarkers()` before DGE analysis, which I understand as a process of merging seperate SCT models.  
Will this process be necessary for Pearson residual normalized data as well, if I want to conduct a DGE analysis?

If there are any other suggestions for regressing out ‘total counts’ before DGE analysis, I would also appreciate that.  
Thank you.