I am planning to integrate single cell dataset of more than 15 samples with each having 2 conditions using Seurat v3. For this, should I analyze (QC, Clustering etc) each dataset seperately before integrating them?
Can anyone suggest the best method to perform this analysis in Seurat?
Thanks Parvathi.
Follow the SCTransform integration vignette on the Seurat website for the preferred workflow. You'll first do some preliminary QC and normalization for each sample individually. You then integrate all of the samples into one object for dimension reduction and clustering.
With that many samples you may want to consider setting only a subset of them as the reference for integration if that makes sense for your data. The reference-based vignette on the same website described it in more detail.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
So in that case, the initial normalization for each sample has to be done using SCTransform. Thank you and sorry for the late reply.
Since my dataset is too large, I am trying the reciprocal PCA for integration. Seurat recommend combining reciprocal PCA with reference-based integration or SCTransform normalization. Will it okay, if I use reciprocal PCA with reference-based integration and SCTransform normalization? My samples are from two conditions. Is it recommended to use the control samples as reference? Or the reference can be any samples?