SIngle cell analysis

Question

[Jeyong]

I'm a beginner in scRNA-seq analysis. I am analyzing CD8 T cell, but I think clustering result is not correct. So I changed FindVariableFeatures(nFeatures = 3000). (default = 2000) Then I can get good clustering result. This means position of CD8 T subclusters are fit to CD8 T cell trajectory in other scRNA-seq T cell study.

My question

1. Can I change parameter 'nFeatures' according to my opinion about clustering result, whenever I want?
2. Could you recommend solution in my situation?

Please tell me about any opinion. Thanks a lot.

Answer 1

[Bastien Hervé]

Yes, if you know what cells you are expecting in your datasets you can play around with the number of variable features, PCs and clustering resolution. Just double check that your clusters have specific gene markers.

From the name of the functions you mentionned, I guess you are using Seurat. You can plot the VariableFeatures using the function of the same name to select number of features specific to your dataset.

[Jeyong]

Hi, Bastien Herve Thanks you for your answer. As you guess, I am using Seurat. I'm sorry I didn't tell about which tool used. The plot function you told may be ().VariableFeaturePlot I checked the plot as you told. Can I decide nfeatures watching standardized variation value(y value) in this plot?

Different solutions you told like PCs and resolution are already used but I can't get result I want. Also I used SCTranform() instead of NormalizeData() function. It didn't give me a result I predict also.

I will decide 'nfeatures' in reference to VariableFeaturePlot. Thanks again.