Bioinformatic characterization of angiotensin-converting enzyme 2, the entry receptor for SARS-CoV-2

Harlan Barker; Seppo Parkkila

Single-cell expression datasets were identified for relevant tissues/cells of trachea (mouse) (37) and lung epithelium (mouse) (38). Using a modified workflow described previously in (39), samples were filtered by Gaussian fit of read count, expressed gene count, and number of cells in which a gene is expressed. Counts were normalized by cell, log transformed, principle component analysis performed with 15 components, and k-nearest neighbors computed using SCANPY (40), and then the full data set normalized with R package 'scran' (41). Batch correction by individual and sample region was performed with SCANPY. The top 1,000 genes with highly differential expression were identified for cluster analysis which was performed with Uniform Manifold Approximation and Projection (UMAP) and force directed graph models. The top 100 marker genes were identified as those with higher expression unique to each cluster by Welch t-test in SCANPY. Expression of the ACE2 gene was mapped onto cluster figures to determine overlap with previously identified cell types or cell type marker genes identified in the literature. Cell type was mapped by expression of known marker genes of cell types expressed in the lung and small intestine, as defined by lists derived from the literature as curated in the CellMarker database (42).