TEMP

Processed standardized expression .h5ad files and differential gene expression (DGE) .h5ad outputs from TEMP datasets whose upstream terms allow redistribution.

This repository contains:

• processed group-replicate bulk or pseudobulk input files under <dataset>/
• DGE outputs under <dataset>/group_rep/ and <dataset>/sep_rep/

The DGE replicate modes indicate whether the DGE analysis was run using all replicates or per replicate (for within-dataset agreement evaluation), shortened to group_rep and sep_rep.

License

The collection curation and packaging are released under CC BY 4.0.

CC BY 4.0 for TEMP curation/packaging; preserve each upstream license, notice, and attribution.

Each upstream dataset retains its original license, attribution, and notice requirements. See upload_manifest.tsv for processed-file provenance and dge_upload_manifest.tsv for per-DGE-file provenance.

Files

Processed Input Files

• l1000_phase1/: l1000_phase1_level3_deg_ready_landmark_processed.h5ad (NCBI/GEO public data; NCBI places no use/distribution restrictions)
• l1000_phase2/: l1000_phase2_level3_deg_ready_landmark_processed.h5ad (NCBI/GEO public data; NCBI places no use/distribution restrictions)
• novartis_batch_2500/: novartis_standardized_processed.h5ad (MIT License)
• op3/: op3_standardized_processed.h5ad (Public GEO data; Open Problems benchmark materials are MIT-licensed)
• sciplex/: srivatsan20_sciplex3_processed.h5ad (CC BY 4.0)
• tahoe/: tahoe_processed.h5ad (CC0 1.0)
• vcpi_0001/: vcpi_0001_standardized_processed.h5ad (CC BY 4.0 for public VCPI releases)
• vcpi_0002/: vcpi_0002_standardized_processed.h5ad (CC BY 4.0 for public VCPI releases)

Differential Expression Files

DGE files are organized as <dataset>/<replicate_mode>/<cell_line>_de.h5ad. Each file contains treatment-vs-control contrasts for one cell line or context.

Dataset	group_rep files	sep_rep files	Total DGE files
l1000_phase1	70	70	140
l1000_phase2	30	30	60
novartis_batch_2500	1	1	2
op3	4	4	8
sciplex	3	3	6
tahoe	48	48	96
vcpi_0001	1	1	2
vcpi_0002	1	1	2
Total	158	158	316

Manifests

• upload_manifest.tsv: one row per processed input file, including path, source, upstream license, and required notice.
• dge_upload_manifest.tsv: one row per DGE file, including dataset, replicate mode, cell line/context, repository path, source, upstream license, and required notice.

AnnData Schema

Processed Input .h5ad

Rows in .obs represent standardized bulk samples or pseudobulk profiles. Core metadata fields include plate, well, cell_type, perturbagen, pert_type, is_control, pert_dose_uM, pert_time_h, suspension_type, tissue, tissue_type, disease, library, stimulation, guide, dataset, assay, development_stage, organism, sex, self_reported_ethnicity, pubchem_cid, psbulk_cells, and psbulk_counts where available.

Columns in .var represent genes and include symbol; some files include additional gene annotations such as merged-gene indicators. .X stores the standardized expression matrix. For single-cell-derived pseudobulk files, .X contains pseudobulk counts and .layers["psbulk_props"] may contain pseudobulk proportions. Bulk-derived datasets may represent raw counts, normalized expression, or platform-specific intensity values depending on the upstream assay and preprocessing.

DGE .h5ad

Rows in .obs represent treatment-vs-control contrasts for one cell line or context. The row index is the contrast label, and metadata is inherited from the treated condition after DGE preprocessing.

Columns in .var represent genes and carry gene annotations from the processed input file.

.X is intentionally absent from DGE files; DGE statistics are stored in .layers:

• logFC
• stdev.unscaled
• stdev.scaled
• CI.L
• CI.R
• AveExpr
• t
• P.Value
• adj.P.Value.within_one_contrast
• adj.P.Value.across_all_contrasts
• B

.uns includes upstream metadata carried from the processed input plus DGE processing metadata such as threshold_filter_cells and qc_filtering_enabled.

Sources And Notices

• l1000_phase1: LINCS L1000 Phase I, GEO GSE92742 - https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92742. Cite LINCS L1000/GSE92742 and preserve upstream attribution.
• l1000_phase2: LINCS L1000 Phase II, GEO GSE70138 - https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70138. Cite LINCS L1000/GSE70138 and preserve upstream attribution.
• novartis_batch_2500: Novartis DRUG-seq U2OS MoABox Dataset - https://zenodo.org/records/14291446. Retain the Novartis/Zenodo MIT license notice and citation.
• op3: OP3 perturbation prediction dataset, GEO GSE279945 - https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE279945. Cite OP3/Open Problems and GSE279945.
• sciplex: scPerturb / SrivatsanTrapnell2020_sciplex3 - https://zenodo.org/records/13350497. Cite scPerturb/Sci-Plex and include CC BY 4.0 attribution.
• tahoe: Tahoe-100M - https://huggingface.co/datasets/tahoebio/Tahoe-100M. Citation requested by the dataset card; attribution still recommended.
• vcpi_0001: Virtual Cell Pharmacology Initiative public release - https://thevirtualcell.com/. Cite VCPI/Ginkgo Datapoints and include CC BY 4.0 attribution.
• vcpi_0002: Virtual Cell Pharmacology Initiative public release - https://thevirtualcell.com/. Cite VCPI/Ginkgo Datapoints and include CC BY 4.0 attribution.