I'm fairly OK with working with gene expression data in Seurat, but it's been a while since I've done anything with Signac and ATAC/multiome single-cell data.
One thing I've noticed recently is that multiome guides for Seurat/Signac all suggest doing QC for nCount_RNA and nCount_ATAC rather than nFeature_RNA and nFeature_ATAC. Following the tutorial for Seurat PBMC expression you see that the advice is to filter on number of features:
pbmc <- subset(pbmc, subset = nFeature_RNA > 200 & nFeature_RNA < 2500 & percent.mt < 5)
However, if you follow the Weighted Nearest Neighbour tutorial for multiome, the advice is to filter on counts:
pbmc <- subset(x = pbmc, subset = nCount_ATAC < 7e4 & nCount_ATAC > 5e3 & nCount_RNA < 25000 & nCount_RNA > 1000 & percent.mt < 20)
Can anyone clarity the discrepancy here?
•
written
by
yura.grabovska •
1 answer
I wish to express my apologies that no Moderator has yet responded to you.
The discrepancy arises from differences in data modalities and quality proxies.
For scRNA-seq alone, nFeature_RNA (unique genes detected) assesses transcriptional complexity: low values flag empty/low-quality droplets; high values suggest doublets. Counts (nCount_RNA) can be inflated by lowly diverse libraries (e.g., stress responses), so features are preferred.
For scATAC-seq (and multiome), nCount_ATAC (total fragments/peaks) directly measures sequencing depth/library complexity, as peak calling depends on counts—nFeature_ATAC correlates strongly with it and adds little unique info. ATAC prioritizes modality-specific QC (TSS enrichment >3–4, nucleosome signal <2–4, %reads in peaks >15–40%) over features.
In multiome, nCount_RNA aligns with ATAC for joint filtering, as RNA capture varies with ATAC depth in 10x kits; thresholds are higher due to deeper sequencing. Visualize VlnPlots/DensityScatters per dataset to tune—e.g., RNA: nCount >1000/<25k; ATAC: >2k/<100k.
Kevin
Log in to answer this question.
-
Could Anyone Help Me with Seurat code
written by ANHAO •
-
batch correction: cds from seurat
written by sooni •
-
Cell type classification using an integrated reference
written by kayah •
-
Converting scMultiome data to loom using SEURAT
written by halimaakhter014 •
-
Cannot RUN PCA using the codes of Cell_Plex_10X_Genomics with Seurat R
written by shweta.johari13 •
-
How to select nFeature parameter in Seurat
written by jinghuang800 •
-
scRNA seq differential gene expression analyses
written by bioinformatics •
-
Qualitative Comparison of 2 scRNAseq analyses
written by ConfusedSheep •
-
Seurat scaling values not between 0 and 1
written by bs58 •
-
Seurat nFeature_RNA filter number in human
written by yueli7
