Knowledge synthesis from 100 million biomedical documents augments the deep expression profiling of coronavirus receptors

AJ Venkatakrishnan; Arjun Puranik; Akash Anand; David Zemmour; Xiang Yao; Xiaoying Wu; Ramakrishna Chilaka; Dariusz K Murakowski; Kristopher Standish; Bharathwaj Raghunathan; Tyler Wagner; Enrique Garcia-Rivera; Hugo Solomon; Abhinav Garg; Rakesh Barve; Anuli Anyanwu-Ofili; Najat Khan; Venky Soundararajan

To systematically profile the transcriptional expression of ACE2 across tissues and cell types, we triangulated single cell RNAseq-based measurements with literature-derived signals to automatically delineate novel, emerging, and known expression patterns ( Figure 2B ; Table S1 ). This approach immediately highlights renal proximal tubular cells and small intestinal enterocytes among the cell types that most robustly express ACE2 (detection in >40% of cells). These cell types are also moderately to strongly associated with ACE2 in the literature. The strong intestinal ACE2 expression is particularly interesting given the emerging clinical reports of fecal shedding and persistence post-recovery which may reflect a fecal-oral transmission pattern 10-12 . Conversely, pancreatic PP cells (gamma cells), pancreatic alpha cells, and keratinocytes show similarly robust ACE2 expression but have not been strongly associated with ACE2 in the literature. This combination suggests either a biological novelty or an experimental artifact. We note that the strong ACE2 expression in pancreatic cell types is derived from only one murine study (Figure S2A 19 ), while ACE2 expression is not observed in gamma or alpha cells from scRNA-seq of human pancreatic islets ( Figure S2B [20] [21] [22] ). While we cannot determine the validity of either observation, this example demonstrates how knowledge synthesis can automatically surface discordant biological signals for further evaluation. author/funder. All rights reserved. No reuse allowed without permission.