scRNA - seq dataset integration
I am currently doing an analysis that involves the integration of multiple scRNA seq samples (13) from different datasets (3). Even though the samples are from the same tissue, the datasets were generated using different sequencing platforms. I have tried integrating it once using the RPCA method in the Seurat workflow, but the resulting clusters were unsatisfactory. Can anyone suggest an appropriate integration workflow and batch correction method for this type of analysis? Thankyou!
• 768 views
•
link
written
by
yadavsoumya647 •
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
scRNA compositional analysis
written by yadavsoumya647 •
-
scRNA datasets integration
written by mrashad
-
Integration of scRNA-seq datasets from different tissues
written by Immreg •
-
scRNAseq normalization
written by sarahmanderni
-
Batch effect correction between Bulk RNA-Seq and scRNA-Seq's clusters
written by ahmad mousavi
-
Batch effect correction methods (Seurat v3, Harmony, fastMNN, Liger)
written by re_raz •
-
scRNA cluster mapping comparision derived from two different analysis
written by tothepoint
-
Merge or Integrate Non-Overlapping Datasets - Different Experiments and Cell Types
written by kebl4660 •
-
What is the right way to merge multiple patient samples in scrna-seq?
written by cogen859 •
-
How to acquire anchor features for multiple dataset single-celll analysis in Seurat 3.0?
written by david.f.stein •
Provide what you have tried. It must be a pipeline or batch correction error
I performed quality control on each dataset separately, according to its original specifications. After that, I SCTransformed each sample individually, then integrated the samples using the Seurat integration workflow. I used the RPCA method for batch correction.
After splitting the UMAP on the basis of dataset, it looked something like this
i am not sure if the batch correction and integration was succesfully done or not. The placement of some clusters are not same in the datasets.
The RPCA is conservative and does less correction, optimal for data coming from the same platforms. Try Harmony or scVI, as they can correct for substantial technical differences.
Thankyou. I'll try the mentioned methods.
I’d recommend plotting UMAP colored by dataset before and after integration to confirm whether clusters are dataset-driven. If RPCA still shows separation by platform, try CCA or Harmony. Successful correction should result in clusters driven by cell type rather than dataset identity
Thankyou. I'll try the mentioned methods.