Control cells
Unperturbed HVG expression profiles and input metadata. No treated expression profiles.
Download controls21.3 MB
New submission
Submit prediction
State-Tahoe-Filtered / c2 / SK-MEL-2
Free dataset
Download inputs
Use these files to generate predictions on your own infrastructure, then upload your prediction below.
.h5ad
Treatment labels
The target conditions your model should predict for this dataset.
HVG feature order
Fixed HVG column positions, not biological gene names. Preserve this order.
Prediction schema
Matrix location, metadata columns, and input conventions for this dataset.
Prepare your prediction
Source documentation- Load the control cells, treatment labels, and ordered HVG features above.
- Generate at least two cells for each requested treatment in its supplied cell context and batch. Keep the supplied feature order.
- Include unperturbed control rows from the supplied inputs under the exact control label below. The evaluator uses these as its baseline.
- Save your predicted expression and condition metadata in an
.h5adfile using the schema below, then select it in the prediction form.
Use the supplied X_hvg representation without reselecting or reordering HVGs. The source does not supply an HVG-to-gene-name mapping.
- Expression matrix
- obsm["X_hvg"]
- Perturbation column
- drugname_drugconc
- Control label
- [('DMSO_TF', 0.0, 'uM')]
- Cell context column
- cell_name
- Batch column
- plate
- Gene set
- STATE HVG / 2,000 fixed feature positions
Keep the supplied feature IDs in uns["hvg_feature_ids"] in the same order as obsm["X_hvg"]. These IDs describe HVG positions, not gene names.
Prepare your prediction
Download the inputs above and follow this dataset’s preparation instructions. Evaluation is not connected for this dataset yet; files will not be uploaded.