analysis of multiple time points * 2 replicates of scRNA-seq about harmony

[ilyakorsunsky]

Dear Bogdan,

You clearly have an interesting experimental design here! I will try to offer advice with respect to Harmony and the general workflow.

If you want a joint embedding of all 8 samples, I recommend that you put all the cells into a single matrix, normalize, scale, PCA, and then Harmonize on whatever conditions you need (e.g. batch, WT/A/B). You describe a few ways of getting all the data into 1 matrix (cell ranger aggr, Seurat Merge). All of these should work and hopefully get you to the same place: a cell by gene raw UMI count matrix.

As you mentioned, you want to retain the different IDs of the 8 experiments for downstream analysis. This information should be stored in Seurat's meta.data dataframe. When you do differential expression analysis, you will use these labels.

I hope that started to answer your questions!

Best,
Ilya

[tanasa]

Thanks, Ilya.