Dataset Viewer
Auto-converted to Parquet Duplicate
loop0/likely-donkey-20_FlowMatching.pkl
dict
loop0/dandy-vortex-20_TargetPredictionModel.pkl
dict
loop0/feasible-lake-25_FlowMatchingWithScore.pkl
dict
loop0/eager-brook-26_FlowMap.pkl
dict
loop1/eager-feather-1_FlowMatching.pkl
dict
loop1/easy-monkey-341_TargetPredictionModel.pkl
dict
loop1/woven-dawn-22_FlowMatchingWithScore.pkl
dict
loop1/copper-thunder-24_FlowMap.pkl
dict
loop2/fresh-bee-21_FlowMatching.pkl
dict
loop2/robust-terrain-39_TargetPredictionModel.pkl
dict
loop2/proud-durian-32_FlowMatchingWithScore.pkl
dict
loop2/easy-spaceship-36_FlowMap.pkl
dict
loop2p5/celestial-fire-40_FlowMatching.pkl
dict
loop2p5/robust-terrain-39_TargetPredictionModel.pkl
dict
loop2p5/soft-brook-35_FlowMatchingWithScore.pkl
dict
loop2p5/ancient-grass-37_FlowMap.pkl
dict
loop3/rich-sunset-44_FlowMatching.pkl
dict
loop3/easy-monkey-341_TargetPredictionModel.pkl
dict
loop3/comic-aardvark-33_FlowMatchingWithScore.pkl
dict
loop3/floral-river-34_FlowMap.pkl
dict
loop4/fast-gorge-13_FlowMatching.pkl
dict
loop4/easy-monkey-341_TargetPredictionModel.pkl
dict
loop4/sunny-jazz-38_FlowMatchingWithScore.pkl
dict
loop4/spring-butterfly-39_FlowMap.pkl
dict
{ "paths_config": "loop0", "config_key": "perturbation_prediction_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/conditional_model/flow_matching/2026-04-26_19-03-48/likely-donkey-20_FlowMatching.pkl", "source_bytes": 113435296170, "dest...
{ "paths_config": "loop0", "config_key": "ct_classifier_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/cell_type_classifier/2026-04-28_16-38-47/dandy-vortex-20_TargetPredictionModel.pkl", "source_bytes": 1082156622, "dest_bytes": 103410...
{ "paths_config": "loop0", "config_key": "prior_flow_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/flow_matching/2026-04-27_10-56-42/feasible-lake-25_FlowMatchingWithScore.pkl", "source_bytes": 5523548808, "dest_bytes": 361228590...
{ "paths_config": "loop0", "config_key": "prior_flow_map_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/flow_map/2026-04-27_12-23-55/eager-brook-26_FlowMap.pkl", "source_bytes": 5524672359, "dest_bytes": 361958965, "ratio": 15.3...
{ "paths_config": "loop1", "config_key": "perturbation_prediction_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/conditional_model/flow_matching/2026-04-02_14-16-39/eager-feather-1_FlowMatching.pkl", "source_bytes": 122572866498, "dest_...
{ "paths_config": "loop1", "config_key": "ct_classifier_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/cell_type_classifier/2026-04-08_00-00-38/easy-monkey-341_TargetPredictionModel.pkl", "source_bytes": 1139827500, "dest_bytes": 116698...
{ "paths_config": "loop1", "config_key": "prior_flow_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/flow_matching/2026-04-04_10-19-42/woven-dawn-22_FlowMatchingWithScore.pkl", "source_bytes": 5971588458, "dest_bytes": 391236705, ...
{ "paths_config": "loop1", "config_key": "prior_flow_map_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/flow_map/2026-04-13_11-57-30/copper-thunder-24_FlowMap.pkl", "source_bytes": 5972711950, "dest_bytes": 391967060, "ratio": 1...
{ "paths_config": "loop2_replicate", "config_key": "perturbation_prediction_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/forward_model/loop2/2026-06-17_17-19-47/fresh-bee-21_FlowMatching.pkl", "source_bytes": 134066877563, "dest_bytes...
{ "paths_config": "loop2_replicate", "config_key": "ct_classifier_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/cell_type_classifier/loop2_replicate/2026-06-25_12-41-22/robust-terrain-39_TargetPredictionModel.pkl", "source_bytes": 215672...
{ "paths_config": "loop2_replicate", "config_key": "prior_flow_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop2_replicate/proud-durian-32_FlowMatchingWithScore.pkl", "source_bytes": 6549915023, "dest_bytes": 429970266, "rati...
{ "paths_config": "loop2_replicate", "config_key": "prior_flow_map_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop2_replicate/easy-spaceship-36_FlowMap.pkl", "source_bytes": 6551038885, "dest_bytes": 430700627, "ratio": 15.2...
{ "paths_config": "loop2p5_replicate", "config_key": "perturbation_prediction_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/forward_model/loop2p5/2026-06-22_12-36-05/celestial-fire-40_FlowMatching.pkl", "source_bytes": 142354736873, "d...
{ "paths_config": "loop2p5_replicate", "config_key": "ct_classifier_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/cell_type_classifier/loop2_replicate/2026-06-25_12-41-22/robust-terrain-39_TargetPredictionModel.pkl", "source_bytes": 2156...
{ "paths_config": "loop2p5_replicate", "config_key": "prior_flow_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop_2p5_replicate/soft-brook-35_FlowMatchingWithScore.pkl", "source_bytes": 6957006523, "dest_bytes": 457235569, "r...
{ "paths_config": "loop2p5_replicate", "config_key": "prior_flow_map_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop_2p5_replicate/ancient-grass-37_FlowMap.pkl", "source_bytes": 6958130388, "dest_bytes": 457965930, "ratio": ...
{ "paths_config": "loop3_replicate", "config_key": "perturbation_prediction_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/forward_model/loop3/2026-06-22_15-10-12/rich-sunset-44_FlowMatching.pkl", "source_bytes": 166627415452, "dest_byt...
{ "paths_config": "loop3_replicate", "config_key": "ct_classifier_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/cell_type_classifier/2026-04-08_00-00-38/easy-monkey-341_TargetPredictionModel.pkl", "source_bytes": 1139827500, "dest_byte...
{ "paths_config": "loop3_replicate", "config_key": "prior_flow_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop3_replicate/comic-aardvark-33_FlowMatchingWithScore.pkl", "source_bytes": 9555862504, "dest_bytes": 520595159, "ra...
{ "paths_config": "loop3_replicate", "config_key": "prior_flow_map_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop3_replicate/floral-river-34_FlowMap.pkl", "source_bytes": 9556986051, "dest_bytes": 521316291, "ratio": 18.3, ...
{ "paths_config": "loop4_replicate", "config_key": "perturbation_prediction_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/forward_model/loop4/2026-07-01_02-07-25/fast-gorge-13_FlowMatching.pkl", "source_bytes": 169587771261, "dest_byte...
{ "paths_config": "loop4_replicate", "config_key": "ct_classifier_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/cell_type_classifier/2026-04-08_00-00-38/easy-monkey-341_TargetPredictionModel.pkl", "source_bytes": 1139827500, "dest_byte...
{ "paths_config": "loop4_replicate", "config_key": "prior_flow_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop4_replicate/sunny-jazz-38_FlowMatchingWithScore.pkl", "source_bytes": 9727387975, "dest_bytes": 530099728, "ratio"...
{ "paths_config": "loop4_replicate", "config_key": "prior_flow_map_path", "source": "/ictstr01/home/icb/alessandro.palma/environment/collab-goettgens-SFC/project_folder/output/checkpoints/prior/loop4_replicate/spring-butterfly-39_FlowMap.pkl", "source_bytes": 9728511530, "dest_bytes": 530820860, "ratio": 18...

LabCompass — Spectral Flow Cytometry haematopoiesis dataset

Measurements underlying LabCompass, a method for generative modeling of experimental design in single-cell data. This dataset contains Spectral Flow Cytometry (SFC) profiles of in vitro haematopoietic differentiation cultures, collected over successive rounds of a closed-loop experimental design cycle.

Each round — a loop — proposes new culture protocols, runs them at the bench, and measures the resulting cells. The measurements from each loop are published here as a separate file.

⚠️ These files are per-loop, not cumulative

loops/loop3.h5ad contains only the cells measured in loop 3 — not loops 0–3 together. Models in the paper are trained on the accumulated data, so a loop's training set is the concatenation of every loop up to and including it:

dataset(N) = concat(dataset(N-1), loopN)

Concatenating them yourself is a few lines of anndata, but the exact chain matters (one loop introduces new protocol axes that must be zero-filled on the earlier data — see below). The reproduction repository ships a script that does it correctly:

git clone https://github.com/theislab/LabCompass.git
python scripts/data/build_loop_datasets.py              # downloads from this repo and builds the chain
python scripts/data/build_loop_datasets.py --variants 500k   # subsampled only: far smaller and faster

Files

Every loop is published in two variants: the full measurement set, and a subsampled version (_500k suffix) intended for fast iteration. The suffix is a naming convention carried over from the source data, not a guaranteed cell count — the subsampled files vary in size.

Loop Full Subsampled Approx. size (full)
0 (baseline) loops/loop0.h5ad loops/loop0_500k.h5ad 36 GB
1 loops/loop1.h5ad loops/loop1_500k.h5ad 2.5 GB
2 loops/loop2.h5ad loops/loop2_500k.h5ad 3.9 GB
2.5 loops/loop2p5.h5ad loops/loop2p5_500k.h5ad 2.7 GB
3 loops/loop3.h5ad loops/loop3_500k.h5ad 6.3 GB
4 loops/loop4.h5ad loops/loop4_500k.h5ad 0.9 GB
4.5 loops/loop4p5.h5ad loops/loop4p5_500k.h5ad 0.5 GB
5 loops/loop5.h5ad loops/loop5_500k.h5ad 6.3 GB

Loop 0 is the baseline screen and is by far the largest. The half-steps (2.5, 4.5) are follow-up rounds within a design cycle and accumulate like any other loop, giving the chain

loop0 → loop1 → loop2 → loop2p5 → loop3 → loop4 → loop4p5 → loop5

The full set is roughly 60 GB; the subsampled set is a few GB.

Format

Each file is an AnnData .h5ad object:

  • X — logicle-transformed SFC intensities: fluorescence channels and morphological scatter features, one row per cell.
  • obs — per-cell metadata, in three groups:
    • Acquisition: experiment_number, experiment_id, replicate, date, well_id, cytometer, cytometer_serial_no, count_beads, cell_counts, source_id.
    • Protocol axes — the culture recipe, and the space LabCompass searches over. Cytokines and small molecules carry their units in the column name, e.g. scf_[ng_ml], tpo_[ng_ml], il3_[ng_ml], gm-csf_[ng_ml], rhflt3l_[ng_ml], ldl_[ng_ml], sr1_[nm], um171_[nm], um729_[µm], butyzamide_[nm], retinoic_acid_[µm], mtg_[µm], 740-yp_[µm], alongside culture conditions such as o2_[%] and hydrogel_type.
    • Annotation: cell-type labels, where available.

experiment_number identifies the physical experiment a cell came from (loop 1, for instance, spans experiments 206–210), which makes it a convenient way to check which loops are present in a concatenated object.

The protocol schema grows across loops

Later loops vary axes that earlier loops never did. Loop 3 introduces il7_[ng_ml], mcsf_[ng_ml] and ly_cocktail_[ul/well], which are absent from loops 0–2.5. When concatenating, these must be zero-filled on the earlier data (they were held at zero, not missing) so both sides share an obs schema. build_loop_datasets.py does this; a naive anndata.concat will silently drop the columns instead.

Loading

import anndata as ad
from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="theislab/LabCompass",
    filename="loops/loop3_500k.h5ad",
    repo_type="dataset",
)
adata = ad.read_h5ad(path)

Citation

@article{labcompass,
  title   = {TODO},
  author  = {Consoli, Lorenzo and Palma, Alessandro and others},
  journal = {TODO},
  year    = {TODO},
}
Downloads last month
48