Hello,
I would like to ask for help with some challenges I am running into during analysis of SMART-seq data.
I have multiple sets of single-cell RNA-seq and single-nuclei RNA-seq data, sequenced with SMART-seq (PLUS kit). I would like to integrate all these data sets.
Normalization
As SMART-seq uses a full-length sequencing protocol, I would like to perform gene length normalization.
TPM normalization method accounts for gene length and sequencing depth. However, by itself, does not fully address the issue of differences in sequencing depth and other technical variations between cells. And of course, is not suitable for cross-sample comparision. Furthermore, use of log(TPM+1) on single-cell/nuclei data with a high abundance of zeros would introduce a bias.
Alternatively, I was considering using sctransform from Seurat package that would enable cross-sample comparision. However, sctransform does not take as input TPM values and therefore I could not normalize for the gene length.
Integration
For the analysis of single nuclei, intronic regions were included in the annotation of each gene to account for the increased abundance of unspliced transcripts. From the initial exploration of the single-nuclei data, I have noticed a significant presence of long non-coding RNAs that I don't see in my single-cell RNA datasets but I am not surprised by this
Is there any good way how to integrate these datasets, please?
I obtained raw gene counts using Cogent NGS Discovery Software pipeline from Takara Bio that manufactures the kit used in this project.
Thanks a lot for considering this!
