single nuclei sequencing questions

Question

[Long]

Good evening, everyone. I hope you all had a great day.

I have recently started to analyze single-nuclei sequencing data and encountered some problems. I hope to receive help from this community.

I have several FASTQ files for single-nuclei sequencing data. To analyze them, I use Cell Ranger 7.1.0, with the mm10 reference file for mice, the same as I use for single-cell sequencing data analysis.

However, when analyzing the data in the same manner as I do for single-cell sequencing, I encountered a problem. After filtering out the 5% of cells with mitochondrial content, there were not many cells left, which led me to suspect that I did something wrong. I referred to the article "Single nucleus RNA-sequencing defines unexpected diversity of cholinergic neuron types in the adult mouse spinal cord." In their analysis, they did not seem to perform the same QC control for mitochondrial cells. My question is: how should we deal with mitochondrial cells in our sample?

Thank you in advance for your support. Long

[LChart]

After filtering out the 5% of cells with mitochondrial content, there were not many cells left, which led me to suspect that I did something wrong.

Do you mean "after excluding cells with more than 5% of UMI from mitochondrial genes"? Can you show a violin plot of % mtRNA?

[Long]

Yes, I can.

[Long]

Oh I see my problem is about the percentage is to much? I should chose 2.5 or even lower?

[LChart]

My guess is that it's not the %Mt filter that's the problem (5 cells appear to be >5%); it's probably the nCount or nFeature threshold. Honestly this does not look like good data - there should be plenty of nuclei with >1000 UMI, and you have very few. Are you certain this is mouse data? Is it a public dataset?

[Long]

This is not public dataset, it from my lab, this is why I hide the identify. Since this is very first single-nuclei sequencing fastq file we have, and this is also my first time handling single-nuclei sequencing data, I just want to know where is the problem. for the features I chose, nfeatures = 2000. I did qc by:

x <- CreateSeuratObject(counts = x, min.features = 200, min.cells = 3, project = 'x')
x$identity <- "x"

x <- NormalizeData(x, verbose = FALSE)
x <- FindVariableFeatures(x, selection.method = "vst", nfeatures = 2000, verbose = FALSE)
x[["percent.mt"]] <- PercentageFeatureSet(x, pattern = "^mt-")

Feature_range <- quantile(x$nFeature_RNA, probs = c(.025, .975))
Count_range <- quantile(x$nCount_RNA, probs = c(.025, .975))

x <- subset(x, subset = nFeature_RNA > Feature_range[1] & 
                     nFeature_RNA < Feature_range[2] & 
                     percent.mt < 5 & # Filtering condition
                     nCount_RNA > Count_range[1] & 
                     nCount_RNA < Count_range[2])

[LChart]

you have min.features=200 in the code block but said "2000".

By the way are these violin plots of the raw data or post-filtering? If not what does the raw data look like? Typically CellRanger will produce a filtered .mtx file for import, I suspect you're not using that file?

[Long]

What I use is this "filtered_feature_bc_matrix". I think I am right?

I mean when I create the Seurat object, I use:

x <- CreateSeuratObject(counts = x, min.features = 200, min.cells = 3, project = 'x'). 

Thanks for your quick response!