Evaluating cell-specific gene expression using single-cell and single-nuclei RNA-sequencing data from human pancreatic islets of the same donors.

Single-cell and single-nuclei RNA-sequencing (scRNA-seq and snRNA-seq) analyze cell-specific transcriptomes. However, only snRNA-seq applies to frozen biobanked samples. For human pancreatic islets, marker genes and reference-based cell type annotation methods are mainly from scRNA-seq datasets and may not be suitable for snRNA-seq. We compared human islet scRNA-seq and snRNA-seq data from the same donors (N = 4) and evaluated annotation methods by studying cell type composition and gene detection, and identified novel marker genes. We compared cell type annotations: (1) manual annotation based on identified marker genes, (2) reference-based annotation using Azimuth's scRNA-seq pancreasref dataset, or (3) Seurat's label transfer from the Human Pancreas Analysis Program (HPAP) scRNA-seq dataset. ScRNA-seq and snRNA-seq identified the same cell types, but predicted cell type proportions differed. Cell type proportion-differences between annotation methods were larger for snRNA-seq. Reference-based annotations generated higher cell type prediction and mapping scores for scRNA-seq than snRNA-seq. Manual annotation identified the novel snRNA-seq markers DOCK10, KIRREL3 (beta cells), STK32B (alpha cells), MECOM, AC007368.1 (acinar cells), LAMC2 and SLC28A3 (ductal cells), which improve snRNA-seq-based annotation. We confirmed ZNF385D as a snRNA-seq beta cell marker and ZNF385D silencing reduced insulin secretion. In conclusion, this study discovered novel snRNA-seq cell type marker genes in human pancreatic islets, and highlights the need for tailored snRNA-seq annotation strategies.