| { |
| "@context": { |
| "@language": "en", |
| "@vocab": "https://schema.org/", |
| "sc": "https://schema.org/", |
| "cr": "http://mlcommons.org/croissant/", |
| "rai": "http://mlcommons.org/croissant/RAI/", |
| "prov": "http://www.w3.org/ns/prov#", |
| "dct": "http://purl.org/dc/terms/", |
| "arrayShape": "cr:arrayShape", |
| "citeAs": "cr:citeAs", |
| "column": "cr:column", |
| "conformsTo": "dct:conformsTo", |
| "data": { |
| "@id": "cr:data", |
| "@type": "@json" |
| }, |
| "dataType": { |
| "@id": "cr:dataType", |
| "@type": "@vocab" |
| }, |
| "equivalentProperty": "cr:equivalentProperty", |
| "examples": { |
| "@id": "cr:examples", |
| "@type": "@json" |
| }, |
| "extract": "cr:extract", |
| "field": "cr:field", |
| "fileObject": "cr:fileObject", |
| "fileProperty": "cr:fileProperty", |
| "fileSet": "cr:fileSet", |
| "format": "cr:format", |
| "includes": "cr:includes", |
| "isArray": "cr:isArray", |
| "isLiveDataset": "cr:isLiveDataset", |
| "jsonPath": "cr:jsonPath", |
| "key": "cr:key", |
| "md5": "cr:md5", |
| "parentField": "cr:parentField", |
| "path": "cr:path", |
| "recordSet": "cr:recordSet", |
| "references": "cr:references", |
| "regex": "cr:regex", |
| "repeated": "cr:repeated", |
| "replace": "cr:replace", |
| "samplingRate": "cr:samplingRate", |
| "separator": "cr:separator", |
| "source": "cr:source", |
| "subField": "cr:subField", |
| "transform": "cr:transform", |
| "value": { |
| "@id": "cr:value", |
| "@type": "@json" |
| } |
| }, |
| "@type": "sc:Dataset", |
| "conformsTo": [ |
| "http://mlcommons.org/croissant/1.1", |
| "http://mlcommons.org/croissant/RAI/1.0" |
| ], |
| "name": "Chem-PerturBridge", |
| "description": "Reviewer-accessible redistributable Chem-PerturBridge release containing standardized processed expression AnnData files and replicate-aware differential gene expression AnnData outputs for eight upstream chemical perturbation transcriptomics datasets.", |
| "url": "https://huggingface.co/datasets/theislab/chem-perturbridge", |
| "version": "1.0.0", |
| "datePublished": "2026-05-04", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html", |
| "citeAs": "Szalata, A., Novitskaia, O., et al. Chem-PerturBridge: a harmonized compendium of small molecule perturbation transcriptomic effects. NeurIPS 2026.", |
| "isLiveDataset": false, |
| "creator": [ |
| { |
| "@type": "Organization", |
| "name": "Helmholtz Munich" |
| }, |
| { |
| "@type": "Organization", |
| "name": "Technical University of Munich" |
| } |
| ], |
| "publisher": { |
| "@type": "Organization", |
| "name": "Theis Lab" |
| }, |
| "prov:wasDerivedFrom": [ |
| { |
| "@type": "prov:Entity", |
| "@id": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92742", |
| "name": "LINCS L1000 Phase I, GEO GSE92742" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70138", |
| "name": "LINCS L1000 Phase II, GEO GSE70138" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://zenodo.org/records/14291446", |
| "name": "Novartis DRUG-seq U2OS MoABox Dataset" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE279945", |
| "name": "OP3 perturbation prediction dataset, GEO GSE279945" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://zenodo.org/records/13350497", |
| "name": "scPerturb / SrivatsanTrapnell2020_sciplex3" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://huggingface.co/datasets/tahoebio/Tahoe-100M", |
| "name": "Tahoe-100M" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://thevirtualcell.com/#vcpi-0001", |
| "name": "Virtual Cell Pharmacology Initiative public release VCPI-0001" |
| }, |
| { |
| "@type": "prov:Entity", |
| "@id": "https://thevirtualcell.com/#vcpi-0002", |
| "name": "Virtual Cell Pharmacology Initiative public release VCPI-0002" |
| } |
| ], |
| "prov:wasGeneratedBy": { |
| "@type": "prov:Activity", |
| "prov:label": "Chem-PerturBridge harmonization and DGE pipeline", |
| "description": "Source expression matrices were standardized into a common AnnData schema, compound identifiers were resolved to PubChem where possible, biological contexts were harmonized, genes were mapped to standardized gene metadata, single-cell datasets were pseudobulked, and differential expression outputs were generated in group-replicate and separate-replicate modes.", |
| "prov:wasAssociatedWith": { |
| "@type": "prov:Agent", |
| "name": "Chem-PerturBridge pipeline" |
| } |
| }, |
| "distribution": [ |
| { |
| "@type": "cr:FileObject", |
| "@id": "readme_md", |
| "name": "README.md", |
| "description": "Human-readable Hugging Face dataset card.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/README.md", |
| "encodingFormat": "text/markdown", |
| "sha256": "d305c75306c142ceb1a832aaeb6cb159e39f6a15df8037e0c986364d41cabd41", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "upload_manifest_tsv", |
| "name": "upload_manifest.tsv", |
| "description": "Processed input file manifest with source, upstream license, and notice fields.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/upload_manifest.tsv", |
| "encodingFormat": "text/tab-separated-values", |
| "sha256": "6f45dacf4824489b65c30b2574f4b4b66afe60f4d3bb05d2ec11085dfc1f9f65", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "dge_upload_manifest_tsv", |
| "name": "dge_upload_manifest.tsv", |
| "description": "DGE output manifest with dataset, replicate mode, cell line or context, repository path, source, upstream license, and notice fields.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/dge_upload_manifest.tsv", |
| "encodingFormat": "text/tab-separated-values", |
| "sha256": "444a33424407167d9bda06b01fd3393f35d10f84d32e9d4d504ccdd344f09a93", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_l1000_phase1", |
| "name": "l1000_phase1_level3_deg_ready_landmark_processed.h5ad", |
| "description": "Processed LINCS L1000 Phase I Level 3 landmark-gene profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/l1000_phase1/l1000_phase1_level3_deg_ready_landmark_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "9c3e3a00da8a84a3305771ffebb0b67a55065ef736f9f51b3b012edb586e7c81", |
| "license": "https://lincsproject.org/LINCS/data/release-policy", |
| "prov:wasDerivedFrom": { |
| "@id": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92742" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_l1000_phase2", |
| "name": "l1000_phase2_level3_deg_ready_landmark_processed.h5ad", |
| "description": "Processed LINCS L1000 Phase II Level 3 landmark-gene profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/l1000_phase2/l1000_phase2_level3_deg_ready_landmark_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "34d198df9eddac5b535a0408794caaf41073fa014c74429d1050e483576fb581", |
| "license": "https://lincsproject.org/LINCS/data/release-policy", |
| "prov:wasDerivedFrom": { |
| "@id": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70138" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_novartis_batch_2500", |
| "name": "novartis_standardized_processed.h5ad", |
| "description": "Processed Novartis DRUG-seq U2OS MoABox profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/novartis_batch_2500/novartis_standardized_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "92856647bce851472cfaa846c5272b643596a8a7b2e3fc2c97cb40ed123b0d49", |
| "license": "https://spdx.org/licenses/MIT.html", |
| "prov:wasDerivedFrom": { |
| "@id": "https://zenodo.org/records/14291446" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_op3", |
| "name": "op3_standardized_processed.h5ad", |
| "description": "Processed OP3 perturbation prediction profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/op3/op3_standardized_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "8eaf7e63adc68029e88184726b8545bc3836c3ef2a25b9185dc4e8b819537e59", |
| "license": [ |
| "https://spdx.org/licenses/CC-BY-4.0.html" |
| ], |
| "prov:wasDerivedFrom": { |
| "@id": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE279945" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_sciplex", |
| "name": "srivatsan20_sciplex3_processed.h5ad", |
| "description": "Processed Sci-Plex3 pseudobulk profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/sciplex/srivatsan20_sciplex3_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "8ad35a20e44bf41cb030a0f063eb6a5727965b8f27d0eda4380bda0e294ccc4c", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html", |
| "prov:wasDerivedFrom": { |
| "@id": "https://zenodo.org/records/13350497" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_tahoe", |
| "name": "tahoe_processed.h5ad", |
| "description": "Processed Tahoe-100M pseudobulk profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/tahoe/tahoe_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "09ac7c8f63a77dc730156c78e361273c3369959640120e5c54241f1a6f82d7b2", |
| "license": "https://spdx.org/licenses/CC0-1.0.html", |
| "prov:wasDerivedFrom": { |
| "@id": "https://huggingface.co/datasets/tahoebio/Tahoe-100M" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_vcpi_0001", |
| "name": "vcpi_0001_standardized_processed.h5ad", |
| "description": "Processed VCPI-0001 public release profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/vcpi_0001/vcpi_0001_standardized_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "5b31cc030e037e89b586403a0d45da80a36f0ba60837e93f782f1c2638f2b227", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html", |
| "prov:wasDerivedFrom": { |
| "@id": "https://thevirtualcell.com/#vcpi-0001" |
| } |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "processed_vcpi_0002", |
| "name": "vcpi_0002_standardized_processed.h5ad", |
| "description": "Processed VCPI-0002 public release profiles standardized for Chem-PerturBridge DGE.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/resolve/main/vcpi_0002/vcpi_0002_standardized_processed.h5ad", |
| "encodingFormat": "application/x-hdf5", |
| "sha256": "edb8bea064daa0f34b94f029be82d1bd2fcce6cae16c7655e313f97e75a36f69", |
| "license": "https://spdx.org/licenses/CC-BY-4.0.html", |
| "prov:wasDerivedFrom": { |
| "@id": "https://thevirtualcell.com/#vcpi-0002" |
| } |
| }, |
| { |
| "@type": "cr:FileSet", |
| "@id": "processed_h5ad_files", |
| "name": "Processed input AnnData files", |
| "description": "All standardized processed input h5ad files included in this release.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/tree/main", |
| "encodingFormat": "application/x-hdf5", |
| "includes": [ |
| "l1000_phase1/*.h5ad", |
| "l1000_phase2/*.h5ad", |
| "novartis_batch_2500/*.h5ad", |
| "op3/*.h5ad", |
| "sciplex/*.h5ad", |
| "tahoe/*.h5ad", |
| "vcpi_0001/*.h5ad", |
| "vcpi_0002/*.h5ad" |
| ] |
| }, |
| { |
| "@type": "cr:FileSet", |
| "@id": "dge_h5ad_files", |
| "name": "All DGE AnnData files", |
| "description": "All differential gene expression h5ad outputs in both replicate modes. Contains 316 files: 158 group_rep and 158 sep_rep.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/tree/main", |
| "encodingFormat": "application/x-hdf5", |
| "includes": [ |
| "l1000_phase1/group_rep/*.h5ad", |
| "l1000_phase1/sep_rep/*.h5ad", |
| "l1000_phase2/group_rep/*.h5ad", |
| "l1000_phase2/sep_rep/*.h5ad", |
| "novartis_batch_2500/group_rep/*.h5ad", |
| "novartis_batch_2500/sep_rep/*.h5ad", |
| "op3/group_rep/*.h5ad", |
| "op3/sep_rep/*.h5ad", |
| "sciplex/group_rep/*.h5ad", |
| "sciplex/sep_rep/*.h5ad", |
| "tahoe/group_rep/*.h5ad", |
| "tahoe/sep_rep/*.h5ad", |
| "vcpi_0001/group_rep/*.h5ad", |
| "vcpi_0001/sep_rep/*.h5ad", |
| "vcpi_0002/group_rep/*.h5ad", |
| "vcpi_0002/sep_rep/*.h5ad" |
| ] |
| }, |
| { |
| "@type": "cr:FileSet", |
| "@id": "dge_group_rep_files", |
| "name": "DGE group_rep AnnData files", |
| "description": "Differential gene expression h5ad outputs generated with all replicates grouped per condition. Contains 158 files.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/tree/main", |
| "encodingFormat": "application/x-hdf5", |
| "includes": [ |
| "l1000_phase1/group_rep/*.h5ad", |
| "l1000_phase2/group_rep/*.h5ad", |
| "novartis_batch_2500/group_rep/*.h5ad", |
| "op3/group_rep/*.h5ad", |
| "sciplex/group_rep/*.h5ad", |
| "tahoe/group_rep/*.h5ad", |
| "vcpi_0001/group_rep/*.h5ad", |
| "vcpi_0002/group_rep/*.h5ad" |
| ] |
| }, |
| { |
| "@type": "cr:FileSet", |
| "@id": "dge_sep_rep_files", |
| "name": "DGE sep_rep AnnData files", |
| "description": "Differential gene expression h5ad outputs generated with replicate-aware separate-replicate conditions. Contains 158 files.", |
| "contentUrl": "https://huggingface.co/datasets/theislab/chem-perturbridge/tree/main", |
| "encodingFormat": "application/x-hdf5", |
| "includes": [ |
| "l1000_phase1/sep_rep/*.h5ad", |
| "l1000_phase2/sep_rep/*.h5ad", |
| "novartis_batch_2500/sep_rep/*.h5ad", |
| "op3/sep_rep/*.h5ad", |
| "sciplex/sep_rep/*.h5ad", |
| "tahoe/sep_rep/*.h5ad", |
| "vcpi_0001/sep_rep/*.h5ad", |
| "vcpi_0002/sep_rep/*.h5ad" |
| ] |
| } |
| ], |
| "recordSet": [ |
| { |
| "@type": "cr:RecordSet", |
| "@id": "processed_observations", |
| "name": "Processed input AnnData observations", |
| "description": "Rows of `.obs` in processed bulk or pseudobulk AnnData files. Each record is one standardized sample/profile used as input to DGE.", |
| "key": { |
| "@id": "processed_observations/sample_id" |
| }, |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/sample_id", |
| "name": "sample_id", |
| "description": "AnnData `.obs` index identifying one processed sample or pseudobulk profile.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/dataset", |
| "name": "dataset", |
| "description": "Source dataset label in AnnData `.obs['dataset']`.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/plate", |
| "name": "plate", |
| "description": "Experimental plate identifier where available.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/well", |
| "name": "well", |
| "description": "Experimental well identifier where available.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/cell_type", |
| "name": "cell_type", |
| "description": "Cell line, cell type, or biological context identifier after standardization.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/perturbagen", |
| "name": "perturbagen", |
| "description": "Small-molecule perturbagen name or source label.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/pubchem_cid", |
| "name": "pubchem_cid", |
| "description": "Resolved PubChem compound identifier where available.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/pert_type", |
| "name": "pert_type", |
| "description": "Perturbation type from source metadata after standardization.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/is_control", |
| "name": "is_control", |
| "description": "Boolean flag indicating control observations.", |
| "dataType": "sc:Boolean", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/pert_dose_uM", |
| "name": "pert_dose_uM", |
| "description": "Perturbation dose converted to micromolar where available.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/pert_time_h", |
| "name": "pert_time_h", |
| "description": "Perturbation exposure time converted to hours where available.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/psbulk_cells", |
| "name": "psbulk_cells", |
| "description": "Number of cells contributing to a pseudobulk profile, or sentinel value for bulk assays where unavailable.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_observations/psbulk_counts", |
| "name": "psbulk_counts", |
| "description": "Total counts contributing to a pseudobulk profile, or sentinel value for bulk assays where unavailable.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| } |
| ] |
| }, |
| { |
| "@type": "cr:RecordSet", |
| "@id": "processed_genes", |
| "name": "Processed input AnnData variables", |
| "description": "Rows of `.var` in processed AnnData files. Each record describes one gene or assayed feature.", |
| "key": { |
| "@id": "processed_genes/gene_id" |
| }, |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "processed_genes/gene_id", |
| "name": "gene_id", |
| "description": "AnnData `.var` index, typically a standardized gene identifier.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_genes/symbol", |
| "name": "symbol", |
| "description": "Gene symbol stored in AnnData `.var['symbol']` where available.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_genes/is_merged", |
| "name": "is_merged", |
| "description": "Optional flag indicating gene records merged during gene metadata harmonization.", |
| "dataType": "sc:Boolean", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| } |
| ] |
| }, |
| { |
| "@type": "cr:RecordSet", |
| "@id": "processed_expression_matrix", |
| "name": "Processed input AnnData expression matrix", |
| "description": "AnnData `.X` matrix in each processed input file. Rows align with `.obs`; columns align with `.var`.", |
| "key": [ |
| { |
| "@id": "processed_observations/sample_id" |
| }, |
| { |
| "@id": "processed_genes/gene_id" |
| } |
| ], |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "processed_expression_matrix/X", |
| "name": "X", |
| "description": "Two-dimensional expression matrix. Single-cell-derived pseudobulk files store pseudobulk counts; bulk-derived files may store raw counts, normalized expression, or platform-specific intensity values depending on the upstream assay.", |
| "dataType": "sc:Float", |
| "isArray": true, |
| "arrayShape": "-1,-1", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "processed_expression_matrix/psbulk_props", |
| "name": "psbulk_props", |
| "description": "Optional AnnData `.layers['psbulk_props']` matrix for pseudobulk proportions.", |
| "dataType": "sc:Float", |
| "isArray": true, |
| "arrayShape": "-1,-1", |
| "source": { |
| "fileSet": { |
| "@id": "processed_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| } |
| ] |
| }, |
| { |
| "@type": "cr:RecordSet", |
| "@id": "dge_contrasts", |
| "name": "DGE AnnData observations", |
| "description": "Rows of `.obs` in DGE AnnData files. Each record is one treatment-vs-control contrast for one cell line or context.", |
| "key": { |
| "@id": "dge_contrasts/contrast" |
| }, |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/contrast", |
| "name": "contrast", |
| "description": "AnnData `.obs` row index formed as treated condition minus control condition.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/replicate_mode", |
| "name": "replicate_mode", |
| "description": "DGE replicate mode: `group_rep` or `sep_rep`.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/cell_line", |
| "name": "cell_line", |
| "description": "Cell line or context encoded by the DGE file name.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/dataset", |
| "name": "dataset", |
| "description": "Source dataset label inherited from the processed input observation metadata.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/perturbagen", |
| "name": "perturbagen", |
| "description": "Treated perturbagen for the contrast.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/pubchem_cid", |
| "name": "pubchem_cid", |
| "description": "Resolved PubChem compound identifier for the treated perturbagen where available.", |
| "dataType": "sc:Integer", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/pert_dose_uM", |
| "name": "pert_dose_uM", |
| "description": "Dose of the treated condition in micromolar.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_contrasts/pert_time_h", |
| "name": "pert_time_h", |
| "description": "Exposure time of the treated condition in hours.", |
| "dataType": "sc:Float", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| } |
| ] |
| }, |
| { |
| "@type": "cr:RecordSet", |
| "@id": "dge_genes", |
| "name": "DGE AnnData variables", |
| "description": "Rows of `.var` in DGE AnnData files. Gene metadata is carried from the processed input files.", |
| "key": { |
| "@id": "dge_genes/gene_id" |
| }, |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "dge_genes/gene_id", |
| "name": "gene_id", |
| "description": "AnnData `.var` index for a gene or assayed feature.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_genes/symbol", |
| "name": "symbol", |
| "description": "Gene symbol from the processed input `.var` metadata where available.", |
| "dataType": "sc:Text", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_genes/is_merged", |
| "name": "is_merged", |
| "description": "Optional merged-gene indicator carried from processed input metadata.", |
| "dataType": "sc:Boolean", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| } |
| ] |
| }, |
| { |
| "@type": "cr:RecordSet", |
| "@id": "dge_statistics", |
| "name": "DGE AnnData statistic layers", |
| "description": "Differential expression statistic matrices stored in AnnData `.layers`. DGE files intentionally omit `.X`; each layer has rows aligned to DGE contrasts and columns aligned to genes.", |
| "key": [ |
| { |
| "@id": "dge_contrasts/contrast" |
| }, |
| { |
| "@id": "dge_genes/gene_id" |
| }, |
| { |
| "@id": "dge_statistics/statistic" |
| } |
| ], |
| "field": [ |
| { |
| "@type": "cr:Field", |
| "@id": "dge_statistics/statistic", |
| "name": "statistic", |
| "description": "Name of the DGE statistic layer.", |
| "dataType": "sc:Text", |
| "isArray": true, |
| "arrayShape": "-1", |
| "value": [ |
| "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" |
| ] |
| }, |
| { |
| "@type": "cr:Field", |
| "@id": "dge_statistics/value_matrix", |
| "name": "value_matrix", |
| "description": "Two-dimensional floating-point matrix for each DGE statistic layer. Rows correspond to contrasts; columns correspond to genes.", |
| "dataType": "sc:Float", |
| "isArray": true, |
| "arrayShape": "-1,-1", |
| "source": { |
| "fileSet": { |
| "@id": "dge_h5ad_files" |
| }, |
| "extract": { |
| "fileProperty": "content" |
| } |
| } |
| } |
| ] |
| } |
| ], |
| "rai:dataCollection": "The public Hugging Face release integrates redistributable processed derivatives from LINCS L1000 Phase I, LINCS L1000 Phase II, Novartis DRUG-seq, OP3, Sci-Plex3, Tahoe-100M, VCPI-0001, and VCPI-0002. Restricted or non-redistributable Chem-PerturBridge project sources are intentionally excluded from this release.", |
| "rai:dataCollectionType": "Secondary reuse of public experimental transcriptomic perturbation datasets with computational harmonization", |
| "rai:dataCollectionMissingData": "Compounds without resolvable PubChem CIDs are retained with source identifiers where possible. Gene coverage varies by assay: L1000 contains landmark genes while RNA-seq and DRUG-seq datasets have broader gene coverage. Not all compounds, doses, times, or contexts appear in all datasets.", |
| "rai:dataCollectionRawData": "Raw or upstream processed data are accessible from the source repositories listed in `prov:wasDerivedFrom` and in the upload manifests. Chem-PerturBridge redistributes only standardized derivatives whose upstream terms allow redistribution.", |
| "rai:dataCollectionTimeframe": "Source datasets were published before this 2026-05-04 release; Chem-PerturBridge harmonization and DGE generation were performed in 2025-2026.", |
| "rai:dataImputationProtocol": "No expression-value imputation is applied. Missing source metadata are retained where possible; compound identifiers may be resolved through automated and manually curated mappings.", |
| "rai:dataPreprocessingProtocol": "Per-dataset preprocessing standardizes observation metadata, compound identifiers, cell or context labels, gene metadata, dose units, and time units. Single-cell datasets are aggregated to pseudobulk profiles. Differential gene expression outputs are generated in group-replicate and separate-replicate modes.", |
| "rai:dataManipulationProtocol": "Compound resolution uses source identifiers, names, InChIKey, SMILES, CAS numbers, cached PubChem lookup, and manual mappings where needed. Doses are converted to micromolar, times to hours, and genes are represented with standardized identifiers and symbols where available.", |
| "rai:dataAnnotationProtocol": "No crowdsourced or task-label annotation was performed. Metadata were derived from source experimental designs and harmonized using automated identifier resolution plus curator-assisted manual mappings for ambiguous or missing compounds, cell contexts, doses, times, and gene metadata where needed.", |
| "rai:dataAnnotationPlatform": "No dedicated crowdsourcing annotation platform was used. Curation was performed by project researchers using source metadata, project scripts, cached lookup results, and manual mapping tables or configuration files.", |
| "rai:dataAnnotationAnalysis": "Automated mappings were checked against public resources such as PubChem, Cellosaurus or source-provided context labels, and gene metadata resources. Manual mappings were used for unresolved or ambiguous cases; no inter-annotator agreement analysis was performed because this was metadata curation rather than independent label annotation.", |
| "rai:annotationsPerItem": "Most metadata fields are inherited from source records or generated deterministically by automated mapping. Manually curated entities generally receive one final harmonized mapping per source entity; repeated independent annotations were not collected.", |
| "rai:annotatorDemographics": "Manual metadata curation was performed by members of the research team.", |
| "rai:machineAnnotationTools": "PubChem lookup and cached/manual compound mappings for compound resolution; Cellosaurus or source context identifiers for cell-line/context standardization; gene metadata harmonization; limma and edgeR for differential expression.", |
| "rai:dataBiases": "The release is biased toward perturbations, assays, cell lines, doses, and time points present in redistributable public source datasets. L1000 and high-throughput screening designs dominate compound coverage, cancer or immortalized cell lines are overrepresented, and assay-specific gene coverage can limit cross-dataset comparisons.", |
| "rai:dataUseCases": "Training and evaluating perturbation models; benchmarking cross-dataset agreement; studying compound and context representations; exploratory drug discovery and transcriptomic perturbation analysis.", |
| "rai:dataLimitations": "The dataset is not suitable for regulatory or clinical toxicity decisions. Cross-assay gene overlap is incomplete, pseudobulk aggregation can obscure single-cell heterogeneity, and differential expression statistics depend on source assay design, control selection, dose, time, and replicate structure. DGE analysis is limited to a single approach and design matrix.", |
| "rai:dataSocialImpact": "The release supports reproducible perturbation modeling and drug discovery research. Results derived from this dataset should be treated as research outputs, not safety certifications or clinical recommendations.", |
| "rai:personalSensitiveInformation": "No identifiable human-subject records or protected health information are included. Some source datasets contain human-derived cell lines or primary-cell/cell-culture contexts, and source metadata may include biological context fields such as tissue, disease, sex, or ethnicity where provided upstream.", |
| "rai:dataReleaseMaintenancePlan": "Hosted on Hugging Face with versioned metadata and manifests. Future updates should preserve upstream attribution, update manifests, and revalidate Croissant metadata before public submission.", |
| "rai:hasSyntheticData": false |
| } |
|
|
