Drug deserts
For each tumor sample, the cosine distance to its single closest drug. Higher = no drug in our 92-drug catalog comes close to reversing this tumor's pathway state.
Per-cohort summary
| Cohort | n samples | median desert score | p95 desert | % in deep desert |
|---|---|---|---|---|
| ccle-hcc@v2024.1 | 22 | 1.093 | 1.420 | 100.0% |
| colon-landscapes@1 | 931 | 0.864 | 1.427 | 63.6% |
| head-neck@1 | 675 | 0.852 | 1.502 | 63.1% |
| breast-umap@18 | 2,284 | 0.845 | 1.449 | 63.2% |
| ccle-aml@v2024.1 | 38 | 0.839 | 1.205 | 73.7% |
| brain-umap@2 | 3,188 | 0.835 | 1.638 | 58.4% |
| mng-umap@100 | 1,298 | 0.833 | 1.490 | 62.9% |
| lung-landscapes@6 | 1,824 | 0.831 | 1.535 | 62.0% |
| melanoma-umap@29 | 1,040 | 0.826 | 1.493 | 60.0% |
| acc-tcga-pancan-2018@2018 | 78 | 0.798 | 1.392 | 64.1% |
| medullo-epn-umap@date-based | 1,358 | 0.797 | 1.554 | 58.3% |
| bladder-landscape@3 | 1,368 | 0.786 | 1.595 | 57.3% |
| ccle-cml@v2024.1 | 14 | 0.678 | 1.203 | 42.9% |
Top 25 most-deserted samples
Even these samples' best-matched drug is barely pointing the right direction. Candidates for first-priority drug discovery.