When to use integration with scRNASeq
1
0
Entering edit mode
15 days ago
Varun Gupta ★ 1.3k

I have a patient sample scRNA-seq data. Pre-treatment and Post Treatment for the same patient. The libraries were prepared on different days, of course because we treated the patient with the drug. The Pre and Post samples were sequenced at the same time in the same run. My question is should I need to perform the integration step (either by harmony or by seurat Integration step) to remove batch effects?

I do want to look at differences in the cells pre and post treatment, so I guess integration makes sense, since it will place similar cells (overlay) from different conditions (pre and post), but I am not sure.

R SEURAT scRNA-seq • 367 views
ADD COMMENT
1
Entering edit mode

With a unique patient, you will not be able to tell if the differences you see are coming from your libraries preparation or your drug effect.

Batch effects correction stand by its name, correcting batches, here you have 1 batch and 2 conditions. You can still try to integrated your 2 conditions, but you will most probably end up masking your drug effect.

ADD REPLY
0
Entering edit mode

Thanks for your reply. Looking at this link from 10x, I do get Pre and Post cluster like this as shown in the link. The 10x capture was done on different days although sequencing run was on the same day and same lane. link

ADD REPLY
0
Entering edit mode

If you are getting the left image, how can you assert the Pre/Post clusters are coming from the preparation or from the treatment ? Sadly, you cannot.

ADD REPLY
1
Entering edit mode
14 days ago
ATpoint 85k

I want to stress here (as I did in many posts before) that methods like harmony do a per-cell batch correction. This means the method aims to position each cell in a UMAP or in a clustering landscape as if there was no batch effect present. It does not do a per-gene correction, so you still cannot do differential expression analysis. It makes sense to do per-cell correction to remove unwanted technical variation. That enables to do clustering and dimensionality reduction between the samples, but you cannot run differential expression. The batch effect is still there and there is no method to maningfully remove this since treatment day and batch are confounded.

ADD COMMENT

Login before adding your answer.

Traffic: 1122 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6