SIngle cell analysis

Question

[Jeyong]

I didn' get any clustering result of scRNA-seq data I expected about many simulations.

I have two different study's data (GSE1(GSM1,2,3,4,5,6), GSE2(GSM1,2,3,4,5,6,7,8)). Data come from prostate cancer. This is my analysis method.

First. Do DoubletFinder : remove doublet for clear clustering result

1. Make seurat objects for each GSM samples. (all GSM)

2. NormalizeData, FindVariableFeatures, ScaleData, RunPCA, FindNeighbors, FindClusters, RunUMAP at each seurat object.

3. Run DoubletFinder and select only 'Singlet'

Second. Merge

1. Use 'merge' function to merge all seurat objects. (10 seurat obejcts)

Third. Run NormalizeData, FindVariableFeatures, ScaleData, RunPCA, FindNeighbors, FindClusters, RunUMAP at merged seurat object.

Fourth. Do Harmony : remove batch effect come from different 2 datasets' environment. (because of 2 data from different 2 studies)

1. Run harmony using GSE_ID

Fifth. Annotation: cell type annotation

1. Use canonical markers that are used in different study. (I use many cell type markers for LE, HE, CE,Endothelial, Fibroblast, Mast, T, B, MNP, ... annotation)

My problem is clustering result is not clear. Clusters have same markers are located in different space at UMAP. And some cell clusters don't have any marker genes. I changed PCA dims(high dims) and harmony variables(GSE_ID, GSM_ID, ...).

I don't know what is the problem in my scRNA-seq analysis process. Please give me any advice. Thank you for reading my question. I'm sorry if there's anything I couldn't be polite because I didn't know English etiquette well. Thanks.