Fig. 2

Detailed analysis of bulk and single cell liver and gut cancers. A Contribution of different signal types to bulk transcriptomes of liver: Relative contribution (y-axis) of different reference single-cell populations (horizontal facets) in explaining bulk transcriptomes (dots) grouped by transcriptome type (x-axis within facets) as indicated by x-axis symbols and labels. The fit was performed with all single-cell reference atlases provided, related groups of reference cell populations indicated by hierarchical labels (top), and the distribution of groups of bulk transcriptomes summarised by their median (horizontal lines) and 1st/3rd quartiles (vertical lines). B Foetal hepatocyte contribution correlated with AFP expression: Foetal hepatocyte contribution to bulk transcriptomes of hepatocellular carcinoma (y-axis) plotted against alpha-fetoprotein (AFP) expression (log10 of TPM, x-axis), for each bulk transcriptome (dots). A best fit linear trend line is shown along with its equation and associated R squared value (p < 2.2 × 10⁻¹⁶, t-test).The fit of the foetal hepatocyte signature to the bulk transcriptome was performed excluding the AFP gene. C UMAP of liver cancer transcriptomes: Dimensionality reduction analysis (UMAP) showing single cell transcriptomes (dots) derived from 1 hepatoblastoma (left) and 8 individual hepatocellular carcinomas (right), grouped by cell type (contours, labels, and colours). Different donors of cancer cells are indicated by different shades of blue. D Similarity of single cell HCC/HB transcriptomes with embryonic and liver reference atlases: Similarity score (logistic regression, colour value) of single-cell transcriptomes of liver cancers (C) grouped by cell type (y-axis) and compared to transcriptomes of reference cell types (x-axis) using a reference consisting of embryonic cells, developmental liver, and post-natal liver. Each rectangle represents a group of cells (indicated by y-axis label) and shows the distribution of similarity scores for those single cells compared to the reference cell population (indicated by x-axis label). E Contribution of different signal types to bulk transcriptomes of the intestines: Relative contribution (y-axis) of different reference single cell populations (horizontal facets) in explaining bulk transcriptomes (dots) grouped by transcriptome type (x-axis within facets) as indicated by x-axis symbols and labels. The fit was performed with all single cell references provided, related groups of reference cell populations indicated by hierarchical labels (top), and the distribution of groups of bulk transcriptomes summarised by their median (horizontal lines) and 1st/3rd quartiles (vertical lines). F UMAP of colorectal carcinoma cell transcriptomes from 25 individuals: Dimensionality reduction analysis (UMAP) showing single-cell transcriptomes (dots) derived from 25 individual HCCs, grouped by cell type (contours, labels, and colours). Different donors of cancer cells are indicated by different shades of blue. G Similarity of single-cell CRC transcriptomes with embryonic and intestine reference atlases: Similarity score (logistic regression, colour value) of single-cell transcriptomes of colorectal cancers (F) grouped by cell type (y-axis) and compared to transcriptomes of reference cell types (x-axis) using a reference atlas consisting of embryonic cells, developmental intestine, and post-natal intestine. Each rectangle represents a group of cells (indicated by the label on the y-axis) and shows the distribution of similarity scores for those single cells compared to the reference cell population (indicated by x-axis label)