Fig 1: Overview of hashing methodology, sample cell-of-origin assignments and pool batch correction, related to Figure 1 and STAR methods(A) Hemotoxylin and Eosin (H&E) staining of intestinal sections demonstrating morphology of samples spanning the times and locations included in transcriptomic atlas (representative images of =3 samples at specified location and similar (+-1pcw) timepoints, each at 20x magnification scale bar=180 µm).(B) Example distribution of B2M mRNA expression in single EPCAM+ cells from an early gestation (8 PCW) sample and late gestation (19 PCW) sample from the same pool (identical sequencing depth and sample preparation conditions) showing reduced B2M mRNA levels in early gestation.(C) Density plot showing the distribution of per cell gene detection rate across different cell compartments. Cells are further broken down into G2M&S Phase cells (dashed line) and G1-phase cells (solid line) based on cluster analysis, as cycling cells in the G2M/S-phase tend to have substantially larger total mRNA content.(D) t-distributed stochastic neighborhood embedding (tSNE) of cells from a representative EPCAM- pool based on their recovered hashing antibody profiles, colored by classification into singlets, doublets or unstained/negative cells following dehashing (STAR methods).(E) tSNE embeddings of EPCAM+ cells from a representative pool, showing embeddings based on TotalSeq antibody tags only (i), in-house labelled antibody tags (ii) and both tags (iii). Cells are colored by sample identities assigned from dual-tag labels. Arrows indicate relevant regions highlighting multiplets (top left, (i) panel); inability to discriminate cells from low gestation samples (Sample 1 and Sample 2) when using TotalSeq tags only (center arrow, (i) panel), while the in-house tag separates cells from Sample 1 (left arrow, (ii) panel) from Sample 2 (right arrow, (ii) panel) while some untagged/negative cells still remain (center arrow, (ii) panel).(F) tSNE overlay comparing TotalSeq (i) and in-house (ii) tag signal in late gestation samples, and early gestation samples (iii-iv) in double-tagged EPCAM+ pool cells. In-house tag signal is stronger than TotalSeq tag for early gestation cells, while being comparable for late gestation cells.(G) tSNE embedding of TotalSeq tag signal in late (i) and early (ii) gestation samples in EPCAM- pools shows similar tag recoveries, in contrast to EPCAM+ pool cells in (F).(H) Expression of compartment markers for stromal cells - THY1/CD90 (i) and epithelial cells - EPCAM (ii) shown as a tSNE overlay over embedding shown in E (iii). Arrows highlight regions of interest, where most unstained/unassigned cells in EPCAM+ pools are either EPCAM- due to poor cell quality or are non-epithelial contaminants expressing stromal or immune markers (center arrow, panels i-ii). Non-epithelial contaminant cell sample-of-origin can be resolved in many cases in double-tagged pools, for instance stromal cells in central region of the embedding in other samples (Figure S1E iii).(I) Example of pool effect batch correction described in STAR methods. A UMAP embedding shows epithelial cells and their hashed pool prior to pool effect correction (i) and post-correction (ii).