I'm annotating clusters from an snRNA-seq dataset from pig samples in Seurat (FindAllMarkers, default Wilcoxon test). The workflow most tutorials show is to filter on p_val_adj < 0.05 and then take the top N genes per cluster by avg_log2FC. I want to sanity-check whether that ordering is actually good enough for annotation, or whether pct.1 and pct.2 should factor into the ranking itself. The canonical markers for celltypes are mostly there where 1/2 genes are there in a cluster while many clusters give the idea of mixed populations. I have tried different PCs and resolutions but not with much success.
1 answer
Cluster annotation in single cell/single nuclei is not rocket science.The Wilcoxon test gives you an idea of your cluster markers but should not be consider for diffenrential analysis. The pct values might giving you some clues whether a gene a representative of a whole cluster or just sporadically expressed in that cluster, one can definitely use this as a cutoff as well.
If your ElbowPlot on PCA has low standard deviation, your UMAP should look like a unique blob, which means that your dataset is quite homogeneous, therefore it is highly probable that your markers will be found across multiple clusters.
Log in to answer this question.
-
Cell-type-specific markers across datasets
written by Chlogst •
-
How to handle a scattered cluster with mixed cell type markers in scRNA-seq subclustering analysis?
written by maplewj •
-
Tutorial: Single-cell RNA-seq: Annotation: Celltype auto annotation with SCSA
written by Julia Ma •
-
WNN in Seurat
written by elengss •
-
Using Seurat function `FindMarkers` to find differentially expressed genes between normal group and…
written by FantasticAI •
I am looking for advice on clustering myself and in my research have seen Seurat be recommended. Would you say that it works for spatial data, e.g. gene expression over time and over multiple tissues?
Please open a new question, your dataset and issues are different from OP's post.
Question is open here: I need help clustering genes across different timepoints AND multiple tissues