Upload README.md with huggingface_hub

README.md

@@ -2,14 +2,19 @@
 license: cc-by-4.0
 tags:
   - biology
-  - proteomics
   - transcriptomics
+  - single-cell
+  - bulk-rna-seq
 ---
 
-# OmicsFM datasets
+# OmicsFM transcriptomics datasets
 
-Expression matrices used to train and evaluate
-[OmicsFM](https://huggingface.co/rednaSander/omicsfm).
+Bulk and single-cell expression matrices used to train and evaluate
+[OmicsFM](https://huggingface.co/rednaSander/omicsfm), prepared as model-ready
+AnnData from public databases.
+
+The proteomics modality lives separately in
+[omicsfm-data-proteomics](https://huggingface.co/datasets/rednaSander/omicsfm-data-proteomics).
 
 ## Naming
 
@@ -22,32 +27,46 @@ what was measured and which identifiers index the features:
 | `_hgnc` | HGNC gene symbols | `TSPAN6` |
 | `_ensembl` | Ensembl gene IDs | `ENSG00000237491` |
 
-The `_uniprot` datasets share one 20,272-accession vocabulary, so a feature
-index means the same protein across proteomics and transcriptomics alike. That
-shared space is what the models consume.
+The `_uniprot` datasets share one 20,272-accession vocabulary with the
+proteomics data, so a feature index means the same protein across every
+modality. That shared space is what the models consume.
 
 The gene-space datasets are the native measurement spaces, kept for comparison
-against methods that operate on genes. Note that bulk and single-cell use
-**different** gene identifiers, so those two are not comparable feature by
-feature; only the `_uniprot` versions are.
+against methods that operate on genes. Bulk and single-cell use **different**
+gene identifiers, so those two are not comparable feature by feature; only the
+`_uniprot` versions are.
 
 ## Contents
 
 | dataset | split | samples | features |
 |---|---|---:|---:|
-| `proteomics_uniprot` | train | 45,188 | 20,272 |
-| `proteomics_uniprot` | valid | 1,568 | 20,272 |
-| `proteomics_uniprot` | test | 2,081 | 20,272 |
+| `bulk_transcriptomics_uniprot` | train | 614,169 | 20,272 |
 | `bulk_transcriptomics_uniprot` | valid | 34,945 | 20,272 |
 | `bulk_transcriptomics_uniprot` | test | 31,102 | 20,272 |
+| `bulk_transcriptomics_hgnc` | train | 614,169 | 67,186 |
 | `bulk_transcriptomics_hgnc` | valid | 34,945 | 67,186 |
 | `bulk_transcriptomics_hgnc` | test | 31,102 | 67,186 |
+| `sc_transcriptomics_uniprot` | train | 3,546,382 | 20,272 |
+| `sc_transcriptomics_uniprot` | valid | 361,037 | 20,272 |
+| `sc_transcriptomics_uniprot` | test | 642,687 | 20,272 |
+| `sc_transcriptomics_ensembl` | train | 3,546,382 | 61,497 |
+| `sc_transcriptomics_ensembl` | valid | 361,037 | 61,497 |
+| `sc_transcriptomics_ensembl` | test | 642,687 | 61,497 |
+
+Splits are grouped so that no study spans a boundary: bulk by ARCHS4 series,
+single-cell by Census dataset. Assignment is deterministic given seed 42.
+
+## Sources
 
-Proteomics splits are by ProteomeXchange accession (`obs.pxd`), so runs from one
-submission never span the train/test boundary.
+| modality | source |
+|---|---|
+| bulk | [ARCHS4](https://archs4.org) human gene-level counts (kallisto, raw estimated counts) |
+| single-cell | [CELLxGENE Census](https://chanzuckerberg.github.io/cellxgene-census/) |
 
-Transcriptomics training splits are not hosted here: they are tens of GB and are
-available on request. The construction scripts are in the repository.
+These are model-ready derivatives, not the primary data: quality filtered,
+harmonised to a controlled vocabulary, and projected onto the shared UniProt
+space. The construction pipeline is in the OmicsFM repository under
+`transcriptomics/`.
 
 ## Use
 
@@ -56,7 +75,7 @@ from huggingface_hub import hf_hub_download
 import anndata as ad
 
 path = hf_hub_download("rednaSander/omicsfm-data",
-                       "bulk_transcriptomics_uniprot/test.h5ad",
+                       "sc_transcriptomics_uniprot/test.h5ad",
                        repo_type="dataset")
 adata = ad.read_h5ad(path)
 ```
@@ -66,4 +85,4 @@ them on first use.
 
 ## Citation
 
-Publication in preparation.
+Publication in preparation. Please also cite ARCHS4 and CELLxGENE Census.