Genetic perturbation representation learning on RPE1, HEPG2, and Jurkat Overall
39.11Average ScoreTxFM-B
Evaluation Results
| Method | Links | |
|---|---|---|
| TxFM-Btrain data=DiverseRNA, # samples=1.4M2026.05 | 39.11 | |
| Ridge PCAtrain data=DiverseRNA, # samples=1.4M2026.05 | 36.74 | |
| STATE-SEtrain data=Tahoe+atlases, # samples=170M2026.05 | 36.43 | |
| (Lib+Log)Normtrain data=n/a2026.05 | 36.29 | |
| TxFM-Btrain data=TF-Sapiens, # samples=57M2026.05 | 35.62 | |
| Cell2Sentencetrain data=Web text+atlases, # samples=50M2026.05 | 35.44 | |
| Linear autoencodertrain data=DiverseRNA, # samples=1.4M2026.05 | 35.38 | |
| Raw datatrain data=n/a2026.05 | 35.27 | |
| (Lib+Log)Norm+5k HVGtrain data=evaluation data2026.05 | 35.08 | |
| AIDO.Cell-100Mtrain data=Atlases, # samples=50M2026.05 | 34.5 | |
| (Lib+Log)Norm+1k HVGtrain data=evaluation data2026.05 | 34.46 | |
| PCAtrain data=DiverseRNA, # samples=1.4M2026.05 | 33.54 | |
| scVI-Ltrain data=DiverseRNA, # samples=1.4M2026.05 | 32.85 | |
| TranscriptFormer-Sapienstrain data=TF-Sapiens, # samples=57M2026.05 | 32.12 | |
| Tahoe-x1-70Mtrain data=Tahoe+atlases, # samples=266M2026.05 | 30.99 | |
| scGPTtrain data=Atlases, # samples=33M2026.05 | 28.96 | |
| scPrint-Ltrain data=Atlases, # samples=50M2026.05 | 27.49 | |
| scCellotrain data=Atlases, # samples=22M2026.05 | 27.04 | |
| UCEtrain data=Atlases, # samples=36M2026.05 | 26.44 | |
| CellPLMtrain data=Tumor atlases, # samples=11M2026.05 | 24.84 | |
| scTabtrain data=Atlases, # samples=22M2026.05 | 24.8 | |
| GeneJEPAtrain data=Tahoe, # samples=100M2026.05 | 23.01 | |
| Geneformer-v2train data=Genecorpus, # samples=104M2026.05 | 19.89 | |
| random label shuffletrain data=n/a2026.05 | 19.43 |