Blame for catalog.json

[
  {
    "algorithm": "AAC.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_aac",
    "description": "Protein representation based on amino acid composition, using the relative frequencies of the 20 standard amino acids in each protein sequence.",
    "id": "probe_aac",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_aac/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 20,
    "n_genes": 18873,
    "name": "AAC",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_aac",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "ALBERT.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.82,
    "dataset_path": "data/embeddings/probe_albert",
    "description": "Protein sequence representation produced by the ALBERT protein language model.",
    "id": "probe_albert",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_albert/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 4096,
    "n_genes": 18453,
    "name": "ALBERT",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_albert",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Google DeepMind AlphaGenome API regulatory predictions. signed log1p per summary feature, z-score across genes, PCA to 256 dimensions.",
    "category": "Regulatory genomics",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.94,
    "dataset_path": "data/embeddings/alphagenome_tss1mb_regulatory_tracksummary_pca256",
    "description": "Gene-level regulatory representation derived from AlphaGenome predictions across multiple regulatory tracks around each transcription start site, summarized and reduced to 256 dimensions.",
    "id": "alphagenome_tss1mb_regulatory_tracksummary_pca256",
    "license": "Original source / license terms",
    "license_url": "https://github.com/google-deepmind/alphagenome",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/alphagenome_tss1mb_regulatory_tracksummary_pca256/metadata.json",
    "modality": "Regulatory genomics",
    "n_dims": 256,
    "n_genes": 20084,
    "name": "AlphaGenome",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/alphagenome_tss1mb_regulatory_tracksummary_pca256",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "APAAC.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.9,
    "dataset_path": "data/embeddings/probe_apaac",
    "description": "Protein representation based on amphiphilic pseudo-amino-acid composition, combining amino-acid composition with sequence-order correlation features.",
    "id": "probe_apaac",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_apaac/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 80,
    "n_genes": 18871,
    "name": "APAAC",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_apaac",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Basenji2 sequence-to-function model (checkpoint: Basenji2 PyTorch port). mean of central 3 output bins around TSS. log1p raw human central3 mean tracks, then dimension-wise z-score across genes.",
    "category": "Regulatory genomics",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 100.0,
    "dataset_path": "data/embeddings/basenji2_tss_131kb_human_central3",
    "description": "Gene-level regulatory representation from Basenji2 predictions on 131-kb DNA windows centered at each transcription start site, summarized across the central output bins.",
    "id": "basenji2_tss_131kb_human_central3",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/14969593",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/basenji2_tss_131kb_human_central3/metadata.json",
    "modality": "Regulatory genomics",
    "n_dims": 5313,
    "n_genes": 20097,
    "name": "Basenji2",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/basenji2_tss_131kb_human_central3",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BERT-BFD.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.82,
    "dataset_path": "data/embeddings/probe_bert_bfd",
    "description": "Protein sequence representation produced by the ProtTrans BERT-BFD model.",
    "id": "probe_bert_bfd",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_bert_bfd/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 1024,
    "n_genes": 18453,
    "name": "BERT-BFD",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_bert_bfd",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BERT-PFAM.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.86,
    "dataset_path": "data/embeddings/probe_bert_pfam",
    "description": "Protein sequence representation produced by a BERT model trained on Pfam protein-domain sequences.",
    "id": "probe_bert_pfam",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_bert_pfam/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 768,
    "n_genes": 18461,
    "name": "BERT-PFAM",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_bert_pfam",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BioBERT-GeneProteinSummary (checkpoint: dmis-lab/biobert-base-cased-v1.2). final hidden layer mean pooling over valid non-special, non-padding tokens.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 95.38,
    "dataset_path": "data/embeddings/biobert_gene_protein_summary",
    "description": "Biomedical-text representation obtained by encoding each gene's NCBI summary plus its UniProt protein summary, when available, with BioBERT and mean-pooling the token representations.",
    "id": "biobert_gene_protein_summary",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/dmis-lab/biobert-base-cased-v1.2",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/biobert_gene_protein_summary/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 768,
    "n_genes": 19168,
    "name": "BioBERT (gene+protein text)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/biobert_gene_protein_summary",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BioConceptVec-CBOW.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 84.79,
    "dataset_path": "data/embeddings/bioconceptvec_cbow",
    "description": "Biomedical concept representation for genes learned from PubMed text with the BioConceptVec CBOW model.",
    "id": "bioconceptvec_cbow",
    "license": "Original source / license terms",
    "license_url": "https://github.com/ncbi/BioConceptVec",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_cbow/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 100,
    "n_genes": 17040,
    "name": "BioConceptVec CBOW",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_cbow",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BioConceptVec-fastText.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 84.79,
    "dataset_path": "data/embeddings/bioconceptvec_fasttext",
    "description": "Biomedical concept representation for genes learned from PubMed text with the BioConceptVec fastText model.",
    "id": "bioconceptvec_fasttext",
    "license": "Original source / license terms",
    "license_url": "https://github.com/ncbi/BioConceptVec",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_fasttext/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 100,
    "n_genes": 17040,
    "name": "BioConceptVec fastText",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_fasttext",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BioConceptVec-GloVe.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 84.79,
    "dataset_path": "data/embeddings/bioconceptvec_glove",
    "description": "Biomedical concept representation for genes learned from PubMed text with the BioConceptVec GloVe model.",
    "id": "bioconceptvec_glove",
    "license": "Original source / license terms",
    "license_url": "https://github.com/ncbi/BioConceptVec",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_glove/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 100,
    "n_genes": 17040,
    "name": "BioConceptVec GloVe",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_glove",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BioConceptVec-SkipGram.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 84.79,
    "dataset_path": "data/embeddings/bioconceptvec_skipgram",
    "description": "Biomedical concept representation for genes learned from PubMed text with the BioConceptVec Skip-Gram model.",
    "id": "bioconceptvec_skipgram",
    "license": "Original source / license terms",
    "license_url": "https://github.com/ncbi/BioConceptVec",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_skipgram/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 100,
    "n_genes": 17040,
    "name": "BioConceptVec Skip-Gram",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/bioconceptvec_skipgram",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "BLAST.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_blast",
    "description": "Protein sequence-similarity representation based on BLAST-derived similarity features against a reference protein set.",
    "id": "probe_blast",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_blast/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 20421,
    "n_genes": 18873,
    "name": "BLAST",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_blast",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "CPC-PROT.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.82,
    "dataset_path": "data/embeddings/probe_cpc_prot",
    "description": "Protein sequence representation produced by CPCProt, a recurrent self-supervised protein model.",
    "id": "probe_cpc_prot",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_cpc_prot/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 512,
    "n_genes": 18453,
    "name": "CPC-PROT",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_cpc_prot",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "deepNF: deep network fusion for protein function prediction.",
    "category": "Networks",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.89,
    "dataset_path": "data/embeddings/deepnf_string_v12_ppmi_svd_fusion",
    "description": "Network representation built from STRING v12 interaction evidence channels using PPMI/SVD features and fusion into a shared low-dimensional gene space.",
    "id": "deepnf_string_v12_ppmi_svd_fusion",
    "license": "Original source / license terms",
    "license_url": "https://string-db.org/cgi/download?species_text=Homo+sapiens",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/deepnf_string_v12_ppmi_svd_fusion/metadata.json",
    "modality": "Networks",
    "n_dims": 256,
    "n_genes": 19071,
    "name": "deepNF (STRING)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/deepnf_string_v12_ppmi_svd_fusion",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "DeepSEA sequence-to-function model (checkpoint: DeepSEA/predict from Kipoi).",
    "category": "Regulatory genomics",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 100.0,
    "dataset_path": "data/embeddings/deepsea_tss_1kb_epigenomic_probs",
    "description": "Regulatory representation from DeepSEA-predicted chromatin and transcription-factor features for 1-kb DNA windows centered at gene transcription start sites.",
    "id": "deepsea_tss_1kb_epigenomic_probs",
    "license": "Original source / license terms",
    "license_url": "https://kipoi.org/models/DeepSEA/predict/",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/deepsea_tss_1kb_epigenomic_probs/metadata.json",
    "modality": "Regulatory genomics",
    "n_dims": 919,
    "n_genes": 20097,
    "name": "DeepSEA",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/deepsea_tss_1kb_epigenomic_probs",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "DepMap CRISPR gene-effect matrix. Gene-effect matrix transposed to genes x models; missing values filled by source gene mean; model/cell-line features standardized across mapped genes; PCA with randomized SVD.",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 92.36,
    "dataset_path": "data/embeddings/depmap_crispr_gene_effect_pca256",
    "description": "Functional gene representation derived from DepMap CRISPR gene-effect profiles across cancer cell lines, standardized and reduced to 256 dimensions with PCA.",
    "id": "depmap_crispr_gene_effect_pca256",
    "license": "Original source / license terms",
    "license_url": "https://depmap.org/portal/download/",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/depmap_crispr_gene_effect_pca256/metadata.json",
    "modality": "Functional",
    "n_dims": 256,
    "n_genes": 18562,
    "name": "DepMap CRISPR",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/depmap_crispr_gene_effect_pca256",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "DNABERT-2 genome language model (checkpoint: DNABERT-2 117M). Mean pooling over final-layer non-special token embeddings. Long CDS sequences split into nucleotide chunks, then length-weighted mean of chunk-level embeddings.",
    "category": "DNA sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.94,
    "dataset_path": "data/embeddings/dnabert2_117m_cds",
    "description": "Gene representation from canonical coding DNA sequences encoded with DNABERT-2, using pooled sequence features to obtain one vector per gene.",
    "id": "dnabert2_117m_cds",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/zhihan1996/DNABERT-2-117M",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/dnabert2_117m_cds/metadata.json",
    "modality": "DNA sequence",
    "n_dims": 768,
    "n_genes": 20085,
    "name": "DNABERT-2",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/dnabert2_117m_cds",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "TransE_l2 knowledge graph embedding",
    "category": "Knowledge graph",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.54,
    "dataset_path": "data/embeddings/drkg_transe",
    "description": "Knowledge-graph gene representation learned with TransE from DRKG’s heterogeneous relations among genes, compounds, diseases, biological processes, and other biomedical entities.",
    "id": "drkg_transe",
    "license": "Apache-2.0 for the DRKG project; underlying source licenses apply",
    "license_note": "DRKG's official source-license table includes CC BY-NC, non-commercial, and mixed-license inputs. Redistribution and use of this derived embedding must comply with all applicable source terms; do not assume unrestricted commercial use.",
    "license_url": "https://github.com/gnn4dr/DRKG/blob/d4bb1974312013c4bd79a13e42c1d9492033f8c7/LICENSE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/313916507a8aea71b3e931b54104cedef19670a6/data/embeddings/drkg_transe/metadata.json",
    "modality": "Knowledge graph",
    "n_dims": 400,
    "n_genes": 18999,
    "name": "DRKG TransE",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/313916507a8aea71b3e931b54104cedef19670a6/data/embeddings/drkg_transe",
    "qc_status": "PASS",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "313916507a8aea71b3e931b54104cedef19670a6",
    "source_url": "https://dgl-data.s3-us-west-2.amazonaws.com/dataset/DRKG/drkg.tar.gz",
    "zero_vectors": 0
  },
  {
    "algorithm": "Enformer sequence-to-function model. mean of central 3 human output bins around TSS. log1p raw human central3 mean tracks, then dimension-wise z-score across genes.",
    "category": "Regulatory genomics",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 100.0,
    "dataset_path": "data/embeddings/enformer_tss_393kb_human_central3",
    "description": "Long-range regulatory representation from Enformer predictions on 393-kb DNA windows centered at gene transcription start sites, summarized across central human output bins.",
    "id": "enformer_tss_393kb_human_central3",
    "license": "Original source / license terms",
    "license_url": "https://tfhub.dev/deepmind/enformer/1",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/enformer_tss_393kb_human_central3/metadata.json",
    "modality": "Regulatory genomics",
    "n_dims": 5313,
    "n_genes": 20097,
    "name": "Enformer",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/enformer_tss_393kb_human_central3",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "ESMB1.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.82,
    "dataset_path": "data/embeddings/probe_esmb1",
    "description": "Protein sequence representation produced by ESM-1b.",
    "id": "probe_esmb1",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_esmb1/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 1280,
    "n_genes": 18453,
    "name": "ESM-1b",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_esmb1",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "ESM2 Transformer (checkpoint: esm2_t33_650M_UR50D, as reported by Zhong). mean pooling over amino-acid token embeddings, as reported by Zhong.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.0,
    "dataset_path": "data/embeddings/esm2",
    "description": "Protein sequence representation from ESM-2 650M, obtained by pooling residue-level embeddings into one vector per human protein/gene.",
    "id": "esm2",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/16764517",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/esm2/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 1280,
    "n_genes": 18690,
    "name": "ESM-2",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/esm2",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "FRoGS-ARCHS4-256 (checkpoint: FRoGS ARCHS4 gene embedding).",
    "category": "Bulk expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 80.93,
    "dataset_path": "data/embeddings/frogs_archs4_256",
    "description": "Expression representation learned by FRoGS from large-scale ARCHS4 bulk RNA-seq data, capturing gene co-expression and transcriptional covariation.",
    "id": "frogs_archs4_256",
    "license": "Original source / license terms",
    "license_url": "https://github.com/chenhcs/FRoGS",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/frogs_archs4_256/metadata.json",
    "modality": "Bulk expression",
    "n_dims": 256,
    "n_genes": 16265,
    "name": "FRoGS (ARCHS4-256)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/frogs_archs4_256",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Gene2Vec.",
    "category": "Bulk expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 89.15,
    "dataset_path": "data/embeddings/gene2vec",
    "description": "Gene co-expression representation learned with Gene2Vec from gene co-occurrence patterns in large-scale expression data.",
    "id": "gene2vec",
    "license": "Original source / license terms",
    "license_url": "https://github.com/jingcheng-du/Gene2vec",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/gene2vec/metadata.json",
    "modality": "Bulk expression",
    "n_dims": 200,
    "n_genes": 17916,
    "name": "Gene2Vec",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/gene2vec",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "GeneCompass Base, 12-layer knowledge-informed cross-species foundation model. none; static learned token embedding weight extracted directly from checkpoint.",
    "category": "Multimodal",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 82.49,
    "dataset_path": "data/embeddings/genecompass_base_gene_token",
    "description": "Multimodal gene-token representation learned from large-scale single-cell transcriptomes together with promoter-sequence, gene-family, co-expression, and gene-regulatory-network priors.",
    "id": "genecompass_base_gene_token",
    "license": "Original source / license terms",
    "license_url": "https://github.com/xCompass-AI/GeneCompass",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/genecompass_base_gene_token/metadata.json",
    "modality": "Multimodal",
    "n_dims": 768,
    "n_genes": 16578,
    "name": "GeneCompass",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/genecompass_base_gene_token",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Geneformer V2 316M input gene-token embedding layer. none; static learned token embedding weight extracted directly from checkpoint.",
    "category": "Single-cell expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.55,
    "dataset_path": "data/embeddings/geneformer_v2_316m",
    "description": "Static gene-token representation extracted from the Geneformer V2 316M input embedding table after pretraining on rank-encoded single-cell transcriptomes.",
    "id": "geneformer_v2_316m",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/ctheodoris/Geneformer",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/geneformer_v2_316m/metadata.json",
    "modality": "Single-cell expression",
    "n_dims": 1152,
    "n_genes": 20007,
    "name": "Geneformer",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/geneformer_v2_316m",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "GenePT-ADA.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 95.03,
    "dataset_path": "data/embeddings/genept_ada",
    "description": "Biomedical-text representation generated from NCBI gene summaries using the GenePT Ada text-embedding model.",
    "id": "genept_ada_conservative",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/10833191",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/genept_ada/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 1536,
    "n_genes": 19099,
    "name": "GenePT Ada",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/genept_ada",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "GenePT-Model3-GeneProtein.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 95.05,
    "dataset_path": "data/embeddings/genept_model3_gene_protein",
    "description": "Biomedical-text representation generated from NCBI gene summaries combined with UniProt protein summaries using GenePT Model 3 embeddings.",
    "id": "genept_model3_gene_protein_conservative",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/10833191",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/genept_model3_gene_protein/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 3072,
    "n_genes": 19103,
    "name": "GenePT Model 3",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/genept_model3_gene_protein",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "binary GO Biological Process profile + TF-IDF + TruncatedSVD-256",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 83.76,
    "dataset_path": "data/embeddings/go_bp_uniprot_tfidf_svd256",
    "description": "Ontology-based representation built from direct, non-negated Gene Ontology Biological Process annotations, weighted with TF-IDF and reduced to 256 dimensions by SVD.",
    "id": "go_bp_uniprot_tfidf_svd256",
    "license": "cc-by-4.0",
    "license_note": "Attribution, release identification, license notice, change notice, source link, and warranty disclaimer are required by CC BY 4.0 and the Gene Ontology citation policy; these are included in the package.",
    "license_url": "https://creativecommons.org/licenses/by/4.0/",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/36c87f4579348d070ba9b1a5fe79f0ac90c46b66/data/embeddings/go_bp_uniprot_tfidf_svd256/metadata.json",
    "modality": "Functional",
    "n_dims": 256,
    "n_genes": 16833,
    "name": "GO TF-IDF SVD-256",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/36c87f4579348d070ba9b1a5fe79f0ac90c46b66/data/embeddings/go_bp_uniprot_tfidf_svd256",
    "qc_status": "PASS",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "36c87f4579348d070ba9b1a5fe79f0ac90c46b66",
    "source_url": "https://release.geneontology.org/2026-08-05/annotations/gaf/HUMAN-uniprot.gaf.gz",
    "zero_vectors": 0
  },
  {
    "algorithm": "GTEx RNA-seq median tissue expression matrix. TPM values were transformed with log1p, then each tissue column was z-scored across mapped genes.",
    "category": "Bulk expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 96.3,
    "dataset_path": "data/embeddings/gtex_tissue_median_log1p_zscore",
    "description": "Expression representation based on median gene-expression profiles across GTEx tissues after log1p transformation and z-score standardization.",
    "id": "gtex_tissue_median_log1p_zscore",
    "license": "Original source / license terms",
    "license_url": "https://gtexportal.org/home/datasets",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/gtex_tissue_median_log1p_zscore/metadata.json",
    "modality": "Bulk expression",
    "n_dims": 68,
    "n_genes": 19353,
    "name": "GTEx",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/gtex_tissue_median_log1p_zscore",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Helix-mRNA (checkpoint: helical-ai/helix-mRNA). final token embeddings from official Helix wrapper, mean pooled over non-special sequence tokens; long transcripts chunked and token-count weighted.",
    "category": "RNA sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.4,
    "dataset_path": "data/embeddings/helix_mrna_full_transcript",
    "description": "mRNA sequence representation obtained by encoding full mature transcripts with Helix-mRNA and pooling transcript-level sequence features into one vector per gene.",
    "id": "helix_mrna_full_transcript",
    "license": "model_license_on_model_card: cc-by-nc-sa-4.0",
    "license_url": "https://huggingface.co/helical-ai/helix-mRNA",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/helix_mrna_full_transcript/metadata.json",
    "modality": "RNA sequence",
    "n_dims": 256,
    "n_genes": 19976,
    "name": "Helix-mRNA",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/helix_mrna_full_transcript",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Hetionet v1.0 heterogeneous biomedical knowledge graph. Built sparse gene x typed-neighbor-feature count matrix; collapsed duplicate gene-feature edges by summation; applied sublinear TF-IDF weighting; reduced with randomized TruncatedSVD; z-scored SVD dimensions across source genes; mapped to master by Entrez ID.",
    "category": "Knowledge graph",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.48,
    "dataset_path": "data/embeddings/hetionet_nonppi_biomedical_context_tfidf_svd256",
    "description": "Knowledge-graph representation built from non-PPI Hetionet gene context, converted to TF-IDF features and reduced to 256 dimensions with SVD.",
    "id": "hetionet_nonppi_biomedical_context_tfidf_svd256",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/268568",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hetionet_nonppi_biomedical_context_tfidf_svd256/metadata.json",
    "modality": "Knowledge graph",
    "n_dims": 256,
    "n_genes": 18787,
    "name": "Hetionet (non-PPI)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hetionet_nonppi_biomedical_context_tfidf_svd256",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "HMMER.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_hmmer",
    "description": "Protein sequence-profile representation based on HMMER-derived profile features.",
    "id": "probe_hmmer",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_hmmer/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 20421,
    "n_genes": 18873,
    "name": "HMMER",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_hmmer",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "HPA subcellular immunofluorescence annotations. Main location, Additional location, and Extracellular location were parsed as semicolon-separated labels. A binary multi-hot location vector was created per gene, duplicate genes were collapsed by max, then location columns were z-scored across mapped genes.",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 67.58,
    "dataset_path": "data/embeddings/hpa_subcellular_location_multihot_zscore",
    "description": "Functional representation from Human Protein Atlas subcellular-localization annotations, encoded as standardized multi-hot cellular-compartment features.",
    "id": "hpa_subcellular_location_multihot_zscore",
    "license": "Original source / license terms",
    "license_url": "https://www.proteinatlas.org/humanproteome/subcellular/data",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hpa_subcellular_location_multihot_zscore/metadata.json",
    "modality": "Functional",
    "n_dims": 50,
    "n_genes": 13581,
    "name": "HPA subcellular",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hpa_subcellular_location_multihot_zscore",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "HPA normal tissue immunohistochemistry annotations. maximum ordinal protein-expression level across annotated cell types within each tissue. HPA Level encoded as Not detected=0, Low=1, Medium=2, High=3. For each gene and tissue, maximum level across annotated cell types was used. Tissue columns were z-scored across mapped genes.",
    "category": "Bulk expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 66.78,
    "dataset_path": "data/embeddings/hpa_normal_ihc_tissue_level_zscore",
    "description": "Functional representation from Human Protein Atlas normal-tissue immunohistochemistry profiles, encoding tissue-level protein expression patterns for each gene.",
    "id": "hpa_normal_ihc_tissue_level_zscore",
    "license": "Original source / license terms",
    "license_url": "https://www.proteinatlas.org/humanproteome/tissue/data",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hpa_normal_ihc_tissue_level_zscore/metadata.json",
    "modality": "Bulk expression",
    "n_dims": 59,
    "n_genes": 13420,
    "name": "HPA tissue",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hpa_normal_ihc_tissue_level_zscore",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "HyenaDNA medium 160k sequence-length checkpoint (checkpoint: LongSafari/hyenadna-medium-160k-seqlen-hf). final hidden-state mean pooling over non-special nucleotide tokens; long CDS split into chunks and token-count weighted.",
    "category": "DNA sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.94,
    "dataset_path": "data/embeddings/hyenadna_medium_160k_cds",
    "description": "Gene representation from canonical coding DNA sequences encoded with the HyenaDNA medium 160k model and pooled into one vector per gene.",
    "id": "hyenadna_medium_160k_cds",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/LongSafari/hyenadna-medium-160k-seqlen-hf",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hyenadna_medium_160k_cds/metadata.json",
    "modality": "DNA sequence",
    "n_dims": 256,
    "n_genes": 20085,
    "name": "HyenaDNA",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/hyenadna_medium_160k_cds",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "LEARNED-VEC.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_learned_vec",
    "description": "Protein sequence representation learned with the unsupervised Learned-Vec method from amino-acid sequence context.",
    "id": "probe_learned_vec",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_learned_vec/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 64,
    "n_genes": 18873,
    "name": "Learned-Vec",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_learned_vec",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Mahi context-aware gene embedding model. mean across all non-global tissue/cell contexts per Entrez gene.",
    "category": "Multimodal",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.13,
    "dataset_path": "data/embeddings/mahi_all_contexts_mean",
    "description": "Multimodal gene representation obtained by averaging Mahi embeddings across tissue and cell contexts, each integrating context-specific network topology with chromatin-accessibility, transcription-factor-binding, histone-modification, and protein-structure information.",
    "id": "mahi_all_contexts_mean",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/datasets/anushaggs/Mahi_gene_embeddings",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/mahi_all_contexts_mean/metadata.json",
    "modality": "Multimodal",
    "n_dims": 3666,
    "n_genes": 18918,
    "name": "Mahi (all-context mean)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/mahi_all_contexts_mean",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Mahi context-aware gene embedding model. global context only.",
    "category": "Multimodal",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.13,
    "dataset_path": "data/embeddings/mahi_global",
    "description": "Multimodal gene representation from the global Mahi context, integrating global network topology with chromatin-accessibility, transcription-factor-binding, histone-modification, and protein-structure information.",
    "id": "mahi_global",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/datasets/anushaggs/Mahi_gene_embeddings",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/mahi_global/metadata.json",
    "modality": "Multimodal",
    "n_dims": 3666,
    "n_genes": 18918,
    "name": "Mahi (global)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/mahi_global",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Mashup-STRING.",
    "category": "Networks",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 81.26,
    "dataset_path": "data/embeddings/mashup_string",
    "description": "PPI/network representation learned with Mashup from STRING interaction networks, preserving integrated network diffusion structure in a low-dimensional gene space.",
    "id": "mashup_string",
    "license": "Original source / license terms",
    "license_url": "https://mashup.csail.mit.edu/",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/mashup_string/metadata.json",
    "modality": "Networks",
    "n_dims": 800,
    "n_genes": 16331,
    "name": "Mashup (STRING)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/mashup_string",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "MSA Transformer (checkpoint: esm_msa1b_t12_100M_UR50S). mean residue representation from layer 12; long proteins split into non-overlapping windows and length-weighted mean pooled.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 96.89,
    "dataset_path": "data/embeddings/msa_transformer_depth1_uniprot_human",
    "description": "Protein sequence representation from the FAIR-ESM MSA Transformer using depth-1 single-sequence inputs, pooled to one vector per human protein/gene.",
    "id": "msa_transformer_depth1_uniprot_human",
    "license": "Original source / license terms",
    "license_url": "https://github.com/facebookresearch/esm",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/msa_transformer_depth1_uniprot_human/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 768,
    "n_genes": 19472,
    "name": "MSA Transformer (depth 1)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/msa_transformer_depth1_uniprot_human",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "MUT2VEC.",
    "category": "Multimodal",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 86.8,
    "dataset_path": "data/embeddings/probe_mut2vec",
    "description": "Multimodal gene representation learned from mutation profiles together with biomedical literature and protein-protein interaction information.",
    "id": "probe_mut2vec",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_mut2vec/metadata.json",
    "modality": "Multimodal",
    "n_dims": 300,
    "n_genes": 17445,
    "name": "Mut2Vec",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_mut2vec",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "NEWT component embedding (checkpoint: not applicable). precomputed component embedding table.",
    "category": "Bulk expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 87.07,
    "dataset_path": "data/embeddings/newt_cellnet",
    "description": "NEWT network representation derived from normalized CellNet tissue- and lineage-specific regulatory-network connectivity scores, encoding each gene’s association with lineage programs.",
    "id": "newt_cellnet",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/16764517",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_cellnet/metadata.json",
    "modality": "Bulk expression",
    "n_dims": 128,
    "n_genes": 17499,
    "name": "NEWT CellNet",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_cellnet",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "NEWT component embedding (checkpoint: not applicable). precomputed component embedding table.",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 68.61,
    "dataset_path": "data/embeddings/newt_go_graph",
    "description": "NEWT component representation derived from the Gene Ontology graph, encoding graph-based functional relationships among genes.",
    "id": "newt_go_graph",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/16764517",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_go_graph/metadata.json",
    "modality": "Functional",
    "n_dims": 256,
    "n_genes": 13788,
    "name": "NEWT GO graph",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_go_graph",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "NEWT component embedding (checkpoint: not applicable). precomputed component embedding table.",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 68.61,
    "dataset_path": "data/embeddings/newt_go_256",
    "description": "NEWT component representation derived from Gene Ontology annotations and provided as a 256-dimensional functional gene embedding.",
    "id": "newt_go_256",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/16764517",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_go_256/metadata.json",
    "modality": "Functional",
    "n_dims": 256,
    "n_genes": 13788,
    "name": "NEWT GO-256",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_go_256",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "NEWT component embedding (checkpoint: not applicable). precomputed component embedding table.",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 76.74,
    "dataset_path": "data/embeddings/newt_msigdb_bundle",
    "description": "NEWT component representation derived from MSigDB pathway and gene-set memberships, encoding shared functional-program context among genes.",
    "id": "newt_msigdb_bundle",
    "license": "Original source / license terms",
    "license_url": "https://zenodo.org/records/16764517",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_msigdb_bundle/metadata.json",
    "modality": "Functional",
    "n_dims": 256,
    "n_genes": 15423,
    "name": "NEWT MSigDB",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/newt_msigdb_bundle",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "node2vec: scalable feature learning for networks.",
    "category": "Networks",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.34,
    "dataset_path": "data/embeddings/node2vec_consensus_ppi",
    "description": "Protein-interaction network representation learned with node2vec on a consensus human protein-protein interaction graph.",
    "id": "node2vec_consensus_ppi",
    "license": "Original source / license terms",
    "license_url": "https://github.com/ylaboratory/gene-embedding-benchmarks",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/node2vec_consensus_ppi/metadata.json",
    "modality": "Networks",
    "n_dims": 128,
    "n_genes": 18356,
    "name": "Node2Vec (PPI)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/node2vec_consensus_ppi",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Nucleotide Transformer v2 (checkpoint: InstaDeepAI/nucleotide-transformer-v2-500m-multi-species). final hidden layer mean pooling over non-special, non-padding tokens; chunk-weighted by token count for long CDS.",
    "category": "DNA sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.94,
    "dataset_path": "data/embeddings/nt_v2_500m_multispecies_cds",
    "description": "Gene representation from canonical coding DNA sequences encoded with Nucleotide Transformer v2 500M multi-species and pooled into one vector per gene.",
    "id": "nt_v2_500m_multispecies_cds",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-500m-multi-species",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/nt_v2_500m_multispecies_cds/metadata.json",
    "modality": "DNA sequence",
    "n_dims": 1024,
    "n_genes": 20085,
    "name": "Nucleotide Transformer v2",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/nt_v2_500m_multispecies_cds",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "OPA2Vec.",
    "category": "Functional",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 92.85,
    "dataset_path": "data/embeddings/opa2vec_goa_human",
    "description": "Ontology representation generated with OPA2Vec from Gene Ontology structure and human GO annotations, embedding genes from their semantic annotation context.",
    "id": "opa2vec_goa_human",
    "license": "Original source / license terms",
    "license_url": "https://github.com/bio-ontology-research-group/opa2vec",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/opa2vec_goa_human/metadata.json",
    "modality": "Functional",
    "n_dims": 200,
    "n_genes": 18660,
    "name": "OPA2Vec",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/opa2vec_goa_human",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Orthrus mature RNA foundation model (checkpoint: Orthrus base 4-track). Orthrus representation() sequence-level embedding. For transcripts longer than max_full_len, length-weighted mean of chunk-level Orthrus embeddings.",
    "category": "RNA sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.4,
    "dataset_path": "data/embeddings/orthrus_base_4track_full_transcript",
    "description": "RNA sequence representation obtained from full mature transcripts with the Orthrus base 4-track model and pooled to one vector per gene.",
    "id": "orthrus_base_4track_full_transcript",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/quietflamingo/orthrus-base-4-track",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/orthrus_base_4track_full_transcript/metadata.json",
    "modality": "RNA sequence",
    "n_dims": 256,
    "n_genes": 19976,
    "name": "Orthrus",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/orthrus_base_4track_full_transcript",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "PrimeKG heterogeneous biomedical knowledge graph. Built sparse gene x typed-neighbor-feature count matrix; collapsed duplicate gene-feature edges by summation; applied sublinear TF-IDF weighting; reduced with randomized TruncatedSVD; z-scored SVD dimensions across source genes; mapped to master by Entrez ID.",
    "category": "Knowledge graph",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 95.62,
    "dataset_path": "data/embeddings/primekg_nonppi_biomedical_context_tfidf_svd256",
    "description": "Knowledge-graph representation built from non-PPI PrimeKG gene context, converted to TF-IDF features and reduced to 256 dimensions with SVD.",
    "id": "primekg_nonppi_biomedical_context_tfidf_svd256",
    "license": "Original source / license terms",
    "license_url": "https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/IXA7BM",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/primekg_nonppi_biomedical_context_tfidf_svd256/metadata.json",
    "modality": "Knowledge graph",
    "n_dims": 256,
    "n_genes": 19216,
    "name": "PrimeKG (non-PPI)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/primekg_nonppi_biomedical_context_tfidf_svd256",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "PROTVEC.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_protvec",
    "description": "Protein sequence representation based on ProtVec, which learns distributed embeddings of amino-acid 3-mers and aggregates them into protein-level vectors.",
    "id": "probe_protvec",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_protvec/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 100,
    "n_genes": 18873,
    "name": "ProtVec",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_protvec",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "PubMedBERT-GeneProteinSummary (checkpoint: microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext). final hidden layer mean pooling over valid non-special, non-padding tokens.",
    "category": "Biomedical text",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 95.38,
    "dataset_path": "data/embeddings/pubmedbert_gene_protein_summary",
    "description": "Biomedical-text representation obtained by encoding each gene's NCBI summary plus its UniProt protein summary, when available, with PubMedBERT and mean-pooling the token representations.",
    "id": "pubmedbert_gene_protein_summary",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/pubmedbert_gene_protein_summary/metadata.json",
    "modality": "Biomedical text",
    "n_dims": 768,
    "n_genes": 19168,
    "name": "PubMedBERT (gene+protein text)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/pubmedbert_gene_protein_summary",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "RNA-FM / rna_fm_t12 RNA language model. Final layer 12 mean token pooling over real RNA tokens only; BOS/EOS excluded; long transcripts split into chunks and chunk means length-weighted.",
    "category": "RNA sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.4,
    "dataset_path": "data/embeddings/rnafm_full_transcript",
    "description": "RNA sequence representation obtained by encoding full mature transcripts with RNA-FM and pooling sequence features into one vector per gene.",
    "id": "rnafm_full_transcript",
    "license": "Original source / license terms",
    "license_url": "https://github.com/ml4bio/RNA-FM",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/rnafm_full_transcript/metadata.json",
    "modality": "RNA sequence",
    "n_dims": 640,
    "n_genes": 19976,
    "name": "RNA-FM",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/rnafm_full_transcript",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "scFoundation / xTrimoGene (checkpoint: models.ckpt). fixed checkpoint embedding table; no cell/context pooling.",
    "category": "Single-cell expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.85,
    "dataset_path": "data/embeddings/scfoundation_gene_pos",
    "description": "Static gene-identity representation extracted from scFoundation’s learned gene-position embedding table after pretraining on large-scale single-cell expression data.",
    "id": "scfoundation_gene_pos",
    "license": "Original source / license terms",
    "license_url": "https://github.com/biomap-research/scFoundation",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scfoundation_gene_pos/metadata.json",
    "modality": "Single-cell expression",
    "n_dims": 768,
    "n_genes": 19062,
    "name": "scFoundation",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scfoundation_gene_pos",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "scGPT pan-cancer checkpoint (checkpoint: scGPT pan-cancer gene-token embedding). none; static learned token embedding weight extracted directly from checkpoint.",
    "category": "Single-cell expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.89,
    "dataset_path": "data/embeddings/scgpt_pancancer",
    "description": "Static gene-token representation extracted from the official scGPT pan-cancer checkpoint, capturing gene identity in a cancer-focused single-cell model.",
    "id": "scgpt_pancancer",
    "license": "Original source / license terms",
    "license_url": "https://github.com/bowang-lab/scGPT",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scgpt_pancancer/metadata.json",
    "modality": "Single-cell expression",
    "n_dims": 512,
    "n_genes": 19071,
    "name": "scGPT (PanCancer)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scgpt_pancancer",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "scGPT whole-human input gene-token embedding layer. none; static learned token embedding weight extracted directly from checkpoint.",
    "category": "Single-cell expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 95.93,
    "dataset_path": "data/embeddings/scgpt_whole_human",
    "description": "Static gene-token representation extracted from the official scGPT whole-human checkpoint pretrained on large-scale single-cell transcriptomic data.",
    "id": "scgpt_whole_human",
    "license": "Original source / license terms",
    "license_url": "https://github.com/bowang-lab/scGPT",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scgpt_whole_human/metadata.json",
    "modality": "Single-cell expression",
    "n_dims": 512,
    "n_genes": 19280,
    "name": "scGPT (Whole)",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scgpt_whole_human",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "scPRINT (checkpoint: medium-v1.5.ckpt). fixed checkpoint embedding table; no cell/context pooling.",
    "category": "Single-cell expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 99.49,
    "dataset_path": "data/embeddings/scprint_medium_v1_5_gene",
    "description": "Learned gene-token representation extracted from scPRINT medium v1.5 after pretraining on large-scale single-cell transcriptomes.",
    "id": "scprint_medium_v1_5_gene",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/jkobject/scPRINT",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scprint_medium_v1_5_gene/metadata.json",
    "modality": "Single-cell expression",
    "n_dims": 256,
    "n_genes": 19995,
    "name": "scPRINT",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/scprint_medium_v1_5_gene",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "SEQVEC.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_seqvec",
    "description": "Protein sequence representation produced by SeqVec, an ELMo-style bidirectional language model trained on protein sequences.",
    "id": "probe_seqvec",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_seqvec/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 1024,
    "n_genes": 18873,
    "name": "SeqVec",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_seqvec",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "SpliceAI splice-site prediction model. aggregate canonical splice-site profiles to one fixed-size vector per gene.",
    "category": "Regulatory genomics",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 100.0,
    "dataset_path": "data/embeddings/spliceai_canonical_splice_site_profile",
    "description": "Splicing representation built from SpliceAI predictions around canonical splice sites and aggregated into a fixed gene-level splice-site profile.",
    "id": "spliceai_canonical_splice_site_profile",
    "license": "Original source / license terms",
    "license_url": "https://github.com/Illumina/SpliceAI",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/spliceai_canonical_splice_site_profile/metadata.json",
    "modality": "Regulatory genomics",
    "n_dims": 606,
    "n_genes": 20097,
    "name": "SpliceAI",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/spliceai_canonical_splice_site_profile",
    "qc_status": "processed_qc_passed",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Residual graph convolutional network link prediction",
    "category": "Networks",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 89.7,
    "dataset_path": "data/embeddings/string_gnn",
    "description": "Network representation learned from the STRING protein-association network with the GNN embedding model used for perturbation prediction.",
    "id": "string_gnn",
    "license": "GenBio AI Community License Agreement",
    "license_url": "https://huggingface.co/datasets/genbio-ai/foundation-models-perturbation/blob/d67ba21019814bb55dbdb448b3b84d18bbc693b8/LICENSE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/f703e997fa18de926e750ff9ec4ad452cbdd8184/data/embeddings/string_gnn/metadata.json",
    "modality": "Networks",
    "n_dims": 256,
    "n_genes": 18028,
    "name": "STRING GNN",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/f703e997fa18de926e750ff9ec4ad452cbdd8184/data/embeddings/string_gnn",
    "qc_status": "PASS",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "f703e997fa18de926e750ff9ec4ad452cbdd8184",
    "source_url": "https://huggingface.co/datasets/genbio-ai/foundation-models-perturbation/blob/d67ba21019814bb55dbdb448b3b84d18bbc693b8/gene_embeddings/GNN_Simple_Official_%28D%3D256%29.h5ad",
    "zero_vectors": 0
  },
  {
    "algorithm": "WaveGC graph spectral wavelet convolution",
    "category": "Networks",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 90.22,
    "dataset_path": "data/embeddings/string_wavegc",
    "description": "Network representation learned from the STRING protein-association network with the WaveGC graph-convolution embedding model used for perturbation prediction.",
    "id": "string_wavegc",
    "license": "GenBio AI Community License Agreement",
    "license_url": "https://huggingface.co/datasets/genbio-ai/foundation-models-perturbation/blob/d67ba21019814bb55dbdb448b3b84d18bbc693b8/LICENSE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/f703e997fa18de926e750ff9ec4ad452cbdd8184/data/embeddings/string_wavegc/metadata.json",
    "modality": "Networks",
    "n_dims": 128,
    "n_genes": 18131,
    "name": "STRING WaveGC",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/f703e997fa18de926e750ff9ec4ad452cbdd8184/data/embeddings/string_wavegc",
    "qc_status": "PASS",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "f703e997fa18de926e750ff9ec4ad452cbdd8184",
    "source_url": "https://huggingface.co/datasets/genbio-ai/foundation-models-perturbation/blob/d67ba21019814bb55dbdb448b3b84d18bbc693b8/gene_embeddings/KG_WaveGC_G_StringDB_combined_0.0_%28D%3D128%29.h5ad",
    "zero_vectors": 0
  },
  {
    "algorithm": "T5.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.82,
    "dataset_path": "data/embeddings/probe_t5",
    "description": "Protein sequence representation produced by the ProtT5 protein language model.",
    "id": "probe_t5",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_t5/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 1024,
    "n_genes": 18453,
    "name": "T5",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_t5",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "TCGA-EMBEDDING.",
    "category": "Bulk expression",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 90.67,
    "dataset_path": "data/embeddings/probe_tcga_embedding",
    "description": "Cancer-expression gene representation learned from TCGA gene-expression profiles, yielding a 50-dimensional vector for each gene.",
    "id": "probe_tcga_embedding",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_tcga_embedding/metadata.json",
    "modality": "Bulk expression",
    "n_dims": 50,
    "n_genes": 18221,
    "name": "TCGA",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_tcga_embedding",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "Universal Cell Embeddings (UCE) (checkpoint: 33l_8ep_1024t_1280.torch). UCE gene encoder projection: Linear(5120->1280), ReLU, LayerNorm.",
    "category": "Multimodal",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 94.95,
    "dataset_path": "data/embeddings/uce_33l",
    "description": "Multimodal gene representation combining ESM2 protein-sequence information with a projection learned by UCE-33L during pretraining on large-scale single-cell transcriptomes.",
    "id": "uce_33l",
    "license": "Original source / license terms",
    "license_url": "https://github.com/snap-stanford/UCE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/uce_33l/metadata.json",
    "modality": "Multimodal",
    "n_dims": 1280,
    "n_genes": 19082,
    "name": "UCE-33L",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/uce_33l",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "UNIREP.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 93.91,
    "dataset_path": "data/embeddings/probe_unirep",
    "description": "Protein sequence representation produced by UniRep, a recurrent neural network trained on large-scale protein sequence data.",
    "id": "probe_unirep",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_unirep/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 5700,
    "n_genes": 18873,
    "name": "UniRep",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_unirep",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  },
  {
    "algorithm": "XLNET.",
    "category": "Protein sequence",
    "category_source": "manual",
    "community_contribution": false,
    "coverage_pct": 91.82,
    "dataset_path": "data/embeddings/probe_xlnet",
    "description": "Protein sequence representation produced by a ProtTrans XLNet protein language model.",
    "id": "probe_xlnet",
    "license": "Original source / license terms",
    "license_url": "https://huggingface.co/spaces/HUBioDataLab/PROBE",
    "metadata_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/blob/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_xlnet/metadata.json",
    "modality": "Protein sequence",
    "n_dims": 1024,
    "n_genes": 18453,
    "name": "XLNet",
    "package_url": "https://huggingface.co/datasets/BoevaLab/Gene-Embedding-Hub-dataset/tree/326b257a20e73df337ece40bb82401e277851f5b/data/embeddings/probe_xlnet",
    "qc_status": "legacy_core_package",
    "source_dataset": "BoevaLab/Gene-Embedding-Hub-dataset",
    "source_revision": "326b257a20e73df337ece40bb82401e277851f5b",
    "zero_vectors": 0
  }
]