OCCAM: Improved Single-Cell Representation Learning and Perturbation Effect Prediction
Paper • 2611.21959 • Published
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Occam learns cell representations and predicts perturbation effects using gene expression, GO, STRING, GenePT, and ESM2 knowledge sources.
checkpoints/occam_small_encoder_decoder.pth.checkpoints/occam_seeds/rhj_fewshot/all_sources/seed_42/
through seed_46/. These are the ABL_N_ALL models used on RHJ, H1, and K562.checkpoints/vcell/norman_combo_NF32ALLA/seed_42/
through seed_46/, for double-perturbation prediction.eval/ and benchmarks/. Knowledge sources: graphs/ and assets/.official_comparisons/, final/, metrics/, and eval_outputs/.Complete virtual-cell checkpoints already contain their encoder and decoder.
Use the selective downloader
to get only the files you need. Checkpoint hashes are in checkpoints/official_manifest.json.
The full stage-two MAE checkpoint is retained under checkpoints/stage2/ for fine-tuning.
Competitor checkpoints are hosted separately. The DiverseRNA pretraining corpus is not hosted here.