I am relatively new to scRNAseq and after successfully analyzing my zebrafish dataset I wish to perform cross-species integration with publicly available datasets from GEO. Using the Seurat V3 integration pipeline after converting mouse and chick gene names to zebrafish orthologs, I'm still finding significant batch effects and it seems like Harmony is a good way to manage these effects. My issue is that the Harmony tutorials and notation online haven't been able to answer two important questions I have which I am getting hung up on.
1) Per the harmony-seurat wrapper notation, I need to create one Seurat object containing each of the three datasets I have:
IMPORTANT DIFFERENCE: In the Seurat integration tutorial, you need to define a Seurat object for each dataset. With Harmony integration, create only one Seurat object with all cells. - https://portals.broadinstitute.org/harmony/SeuratV3.html
Can I create this single object by merging seurat objects for each dataset? If no, does anyone have an idea of how to do this?
Below is my approach:
fish.data <- Read10X("fish path/", gene.column = 4, unique.features = TRUE)
mouse.data <- Read10X("mouse path/", gene.column = 4, unique.features = TRUE)
chick.data <- Read10X("chick path/", gene.column = 4, unique.features = TRUE)
## added additional meta.data to fish, code not shown
fish <- AddMetaData(fish_WT, metadata="zfish", col.name="species")
chick <-AddMetaData(chick, metadata="chick", col.name="species")
mouse <-AddMetaData(mouse1, metadata="mouse", col.name="species")
## merging into one seurat object ##
FCM.merged <- merge(fish, y = c(chick, mouse), add.cell.ids = c("ZF", "CHICK", "MOUSE"), project = "FCM_Integrated")
FCM.merged
## run harmony, follow with standard notation https://portals.broadinstitute.org/harmony/SeuratV3.html ##
RunHarmony("species", plot_convergence = TRUE)
2) If I create a subset of cell types from the soon to be integrated dataset, would I be able to / have to use RunHarmony again on the subset in lieu of RunPCA? How does it work if you run Harmony on a dataset that is already integrated? 2.1) Followup question, what would happen if I ran Harmony on the integrated dataset I previously made using Seurat V3?
Thank you very much, any help or direction to resources for a novice single-cell RNA seq user would be fantastic.
S.
