Non-neural expression of SARS-CoV-2 entry genes in the olfactory epithelium suggests mechanisms underlying anosmia in COVID-19 patients

David Brann; Tatsuya Tsukahara; Caleb Weinreb; Darren W. Logan; Sandeep Robert Datta

Analysis of WOM scSeq data were performed in python using the open-source Scanpy software starting from the raw UMI count matrix of the 40179 cells passing the initial filtering and QC criteria described above. UMIs were total-count normalized and scaled by 10,000 (TPT, tag per ten-thousands) and then log-normalized. For each gene, the residuals from linear regression models using the total number of UMIs per cell as predictors were then scaled via z-scoring. PCA was then performed on a set of highlyvariable genes (excluding OR genes) calculated using the "highly_variable_genes" function with parameters: min_mean=0.01, max_mean=10, min_disp=0.5. A batch corrected neighborhood graph was constructed by the "bbknn" function with 42 PCs with the parameters: local_connectivity=1.5, and embedding two-dimensions using the UMAP function with default parameters (min_dist = 0.5). Cells were clustered using the neighborhood graph via the Leiden algorithm (resolution = 1.2). Identified clusters were manually merged and annotated based on known marker gene expression. We The filtered HBC lineage dataset containing 21722 cells was analyzing in python and processed for visualization using pipelines in SPRING and Scanpy (66, 67) . In brief, total counts were normalized to the median total counts for each cell and highly variable genes were selected using the SPRING gene filtering function ("filter_genes") using parameters (90, 3, 3) . The dimensionality of the data was reduced to 20 using principal components analysis (PCA) and visualized in two-dimensions using the UMAP method author/funder. All rights reserved. No reuse allowed without permission.