“…Integrative analyses of such large-scale datasets originating from various samples, different platforms and different institutions globally, offer unprecedented opportunities to establish a comprehensive picture of cell landscape. To this end, various community generated large-scale atlas-level single cell reference data, such as the Human Cell Atlas (HCA 6 ), Human Tumor Atlas Network 7 , BRAIN Initiative Cell Census Network 8 , Human Lung Atlas 9 , Human Gut Atlas 10 , Human BioMolecular Atlas Program (HuBMAP 11 ), The Tabula Sapiens 12 , hECA 13 etc., and recently great achievements has been made in the building of pan-tissue single-cell transcriptome atlases covering more than a million cells, including 500 cell types,across more than 30 human tissues from 68 donors 12,[14][15][16][17] .These references data facilitate the automatically cell type annotations in a supervised way without prior marker gene annotations [18][19][20][21][22][23][24] . It is obviously that integrating more reference datasets or combining these atlas-level data will improve the cell type annotations 18,19,23,25 , and various integration methods for single cell annotations have been presented 18,26,27,28 .…”