Ten Genes
Her mouse findings, next to the human evidence.
At Emory Winship, with Jason Yoo and Dr. Richard Moffitt, Ridhi Koner analysed mouse single-cell RNA-seq from Burdziak et al., Science 2023 (GEO GSE207938) and found ten genes elevated in pancreatic cancer (PDAC) compared with precancerous PanIN lesions. This page maps each mouse gene to its human ortholog and asks a narrow question: does public human pancreatic cancer data point the same way?
The agreement view
Each row is one gene. The first bar is her mouse result (log2 fold change, PDAC vs PanIN). The second is the human bulk comparison (log2 difference in median expression, TCGA pancreatic tumors vs GTEx normal pancreas). Both are log2 ratios, but they compare different things, so read them as direction, not as the same measurement. Select a row to open that gene.
One gene at a time
Caveats
Read every agreement call above through these four limits.
Mouse is not human
Her data come from engineered mice (KC and KPC models driven by mutant Kras, with one p53 allele removed in KPC). A shared direction in human tumors is reassuring, but mouse and human genes are regulated differently, and an ortholog can play a different role.
Bulk tumor vs sorted epithelial cells
Her cells were sorted mKate2+ epithelial cells. The human TCGA samples are bulk tumor pieces that also hold fibroblasts, immune cells and blood vessels, and normal GTEx pancreas is mostly acinar tissue. A gene can rise in bulk just because the tissue mix changed. That is why the agreement rule asks for cell-level human evidence.
One PanIN mouse
The PanIN side (K4) has 1,254 cells from a single mouse, sequenced in two runs. Thousands of cells do not replace biological replicates, so her p values describe cells, not animals. Treat the mouse result as exploratory.
Correlation, not causation
Every number here is an association: higher expression alongside cancer, or alongside shorter survival. None of it shows that a gene drives PDAC, and survival links in public cohorts often fail to replicate, as TM4SF1 shows.
How the verdict is decided
A different comparison is also worth keeping in mind: she compared PDAC with PanIN, while the human bulk data compare PDAC with normal pancreas. Only the S100A6 studies include human PanIN lesions.
Sources and provenance
Every number on this page was precomputed into /ten-genes/data.json, which carries the block below. Where a value could not be retrieved, the page says "not available" instead of estimating it.