Before you can model a cell, you have to know what kind of cell it is. Today’s digest follows the quiet revolution that turned cell-typing from a manual art into an automated, reference-driven science — the labeling layer beneath the virtual cell. Regev et al.’s Human Cell Atlas set the goal: ’the time is ripe to complete the 150-year-old effort to identify all cell types in the human body.’ The Tabula Sapiens Consortium built the reference — ’nearly 500,000 cells from 24 different tissues,’ more than ‘400 cell types.’ Aran et al.’s SingleR gave ‘a novel computational framework for the annotation of scRNA-seq by reference,’ and used it to find a real profibrotic macrophage. Domínguez Conde et al.’s CellTypist is ‘a machine learning tool for rapid and precise cell type annotation.’ Xu et al.’s scANVI learns ’to automatically assign cell type labels in a new dataset based on existing annotations.’ And Lotfollahi et al.’s scArches maps new data onto a reference ‘without sharing raw data.’ A common language for cells.
