Atlas
Sign in
Models / single-cell / gears / benchmarks

single-cell/gears

Visibility: PUBLIC
Model cardFilesBenchmarks
Group by

Other Benchmarks drawn from no hub dataset you can see

Perturbation response (Pearson on Δ expression) single-cell/perturbation-response-delta-pearson

atlas-full-v1

Metric Value 95% CI n Rank Best in catalogue Source Overlap Board
Pearson Δ, all genes (trained) ↑ higher is better · supervised
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.575 [0.508, 0.637]43
Two-gene combination, both genes seen in training0.589 [0.498, 0.677]19
Single gene, unseen in training0.460 [0.403, 0.511]36
 0.543 best [0.506, 0.578] 107 1 / 2 0.543 reproduced none declared Board
Pearson Δ, top-20 DE genes (trained) ↑ higher is better · supervised
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.793 [0.694, 0.870]43
Two-gene combination, both genes seen in training0.798 [0.714, 0.883]19
Single gene, unseen in training0.641 [0.524, 0.747]36
 0.750 best [0.695, 0.804] 107 1 / 2 0.750 reproduced none declared Board
MSE, top-20 DE genes (trained) ↓ lower is better · log1p-normalised expression² · supervised
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.229 [0.182, 0.276]43
Two-gene combination, both genes seen in training0.200 [0.150, 0.253]19
Single gene, unseen in training0.207 [0.161, 0.251]36
 0.214 best [0.185, 0.241] 107 1 / 2 0.214 reproduced none declared Board
Direction accuracy, top-20 DE genes (trained) ↑ higher is better · supervised
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.870 [0.814, 0.914]43
Two-gene combination, both genes seen in training0.882 [0.805, 0.939]19
Single gene, unseen in training0.796 [0.736, 0.850]36
 0.849 best [0.818, 0.879] 107 1 / 2 0.849 reproduced none declared Board
MSE, all genes (trained) ↓ lower is better · log1p-normalised expression² · supervised · secondary
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.007 [0.006, 0.008]43
Two-gene combination, both genes seen in training0.006 [0.005, 0.007]19
Single gene, unseen in training0.004 [0.003, 0.004]36
 0.006 best [0.005, 0.006] 107 — 0.006 reproduced none declared Board
Discrimination score (trained) ↑ higher is better · supervised · secondary
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.794 [0.724, 0.864]43
Two-gene combination, both genes seen in training0.825 [0.737, 0.903]19
Single gene, unseen in training0.623 [0.532, 0.705]36
 0.764 best [0.720, 0.806] 107 — 0.764 reproduced none declared Board
DE-gene overlap, top 20, pseudobulk proxy (trained) ↑ higher is better · supervised · secondary
By subgroup: 4 levels
subgroupValue [95% CI]n
Two-gene combination, neither gene seen in training—fewer than 10 items9
Two-gene combination, one gene seen in training0.257 [0.215, 0.300]43
Two-gene combination, both genes seen in training0.300 [0.234, 0.371]19
Single gene, unseen in training0.175 [0.143, 0.206]36
 0.238 best [0.216, 0.261] 107 — 0.238 reproduced none declared Board