TCR-DeepInsight Reference Datasets

Reference single-cell TCR immune profiling datasets for the TCR-DeepInsight analysis. We have excluded two controlled-access datasets (1) AML dataset from Abbas et al. 2021 (EGAS00001004894) and (2) Kawasaki disease dataset from Wang et al., 2021 (OEP001162).

• human_gex_reference_v2.h5ad: Processed H5AD file for transcriptome features from the single-cell TCR immune profiling datasets. anndata.read_h5ad
• human_tcr_reference_v2.h5ad: Processed H5AD file for unique TCR clonotypes single-cell TCR immune profiling dataset. anndata.read_h5ad
• Yi_2023_Ankylosing_Spondylitis.h5ad: Processed H5AD file from Yi et al., 2023 (GSE216885). anndata.read_h5ad
• GSE272993_cd8_nn_labeled_FINAL.fl_tcr.match_v2_5.transfered.h5ad: Processed H5AD file from Wang et al., 2024 (GSE272993). anndata.read_h5ad
  
  
• human_bulk_tcr_reference.parquet: PARQUET file for bulk TCRβ sequencing. pandas.read_parquet
  • human_bulk_tcr_reference.cd4.parquet: PARQUET file for bulk TCRβ sequencing for CD4-sorted T cells
    human_bulk_tcr_reference.mait.parquet: PARQUET file for bulk TCRβ sequencing for MAIT cells
    human_bulk_tcr_reference.treg.parquet: PARQUET file for bulk TCRβ sequencing for Treg cells
    human_bulk_tcr_reference.cd8.parquet: PARQUET file for bulk TCRβ sequencing for CD8-sorted T cells
    
    
• Yi_2023_Ankylosing_Spondylitis.h5ad. anndata.read_h5ad
  Processed dataset from K. Yi et al. Analysis of Single‐Cell Transcriptome and Surface Protein Expression in Ankylosing Spondylitis Identifies OX40 ‐Positive and Glucocorticoid‐Induced Tumor Necrosis Factor Receptor–Positive Pathogenic Th17 Cells. Arthritis & Rheumatology 75, 1176–1186 (2023).
• GSE272993_cd8_nn_labeled_FINAL.fl_tcr.match_v2_5.transfered.h5ad. anndata.read_h5ad
  Processed dataset from K. Wang et al. Combination anti-PD-1 and anti-CTLA-4 therapy generates waves of clonal responses that include progenitor-exhausted CD8+ T cells. Cancer Cell, S1535610824003064 (2024).

• human_gex_reference_v2.scatlasvae.ckpt: pretrained weight state dict from scAtlasVAE model for human_gex_reference_v2.h5ad. torch.load
• human_bert_pseudosequence.tcr_v2.ckpt: pretrained weight state dict from BERT for human_tcr_reference_v2.h5ad. torch.load
  
• human_bert_pseudosequence_pca.tcr_v2.pkl: pretrained PCA weight for human_tcr_reference_v2.h5ad. pickle.load