Hello everyone,
I have a cohort of ALS and normal control of spinal cord (10 ALS spinal cord lumbar+cervical & 10 control spinal cord same regions). Because of receiving control spinal cords late I did not randomized the case/control sample; 10 ALS were processed using chemistry version 3 and 10 control+1 ALS were processed using chemistry version 3.1 (one of the ALS samples were reprocessed, it means that once it processed with v3 and once with v3.1 but I did use the latter for analysis as I didn't want to processed all control separate from ALS). So, I'm aware that it could have significant chemistry effect. What I did is to integrate the dataset as following: 9 ALS v.3+ 1ALS v.3+10 control v 3.1+1ALS v3.1 (one of the ALS samples were processed two times with two different chemistry). To see if chemistry has significant effect I integrated the aforementioned dataset to see if the ALS replicates which was processed 2 times are mixed together. Below is the UMAP plot of data which are 9 ALS+10 control, ALS replicate chemistry v3.1, and same ALS replicate v3 respectively from left hand side.
I don't see major difference between the second and third UMAP plot. I was wondering should I ignore the chemistry effect with that much difference?
If yes, should stop regressing out the chemistry effect to I still need to do so?
Finally, is that a reliable way to test the chemistry effect?
I appreciate any input on this.
Paria
You should generate the UMAP without integration and see if batch effects are present. If so, then it suggests that you need integration and there is an effect from the chemistry. Otherwise I would conclude that chemistry does not have a significant effect.
Thanks for your reply. I did pseudo bulk analysis and there is chemistry effect. However, the same plot before correction is as below. In the right hand side plot the green dots are hiding behind other colors.