scRNA-seq: To merge technical sequencing replicates or not to merge?

Question

[Kanai]

HI all,

I have sorted single cells in 10 96well plates and performed SmartSeq2. A couple of the plates had to be resequenced as the desired depth of 1Mreads/cell was not reached for a few of them. I am wondering if in my analysis I should merge the counts of the replicates for each cell or if I should treat them as separate. Here's a tSNE plot of the normalized data:

My question is : Should I merge the counts from 1 and 1R (and also 3,3R and 5,5R) in my count matrix or would that not change anything and I should just leave them as is?

Conceptually merging technical replicates for bulkRNA if they would cluster like this would kind of make sense, but I don't know much about single-cell RNA and what would be a sound approach here. Any help would be greatly appreciated.

Thanks!

Answer 1

[swbarnes2]

If you sequence the same library a few times, you should combine the fastqs together from the beginning so UMIs are handled properly, the same way you'd combine the fastqs from different lanes of the same sequencing run.

[Kanai]

Great, thanks so much for the advice!