Gradient Biotech

TCR/BCR repertoire

The repertoire_analysis pipeline is the canonical Immunology engine for TCR/BCR repertoire methods. Oncology calls this shared engine for repertoire summaries.

Inputs

Supported tabular inputs include:

  • 10x Immune Profiling contig annotations
  • AIRR-style fields
  • bulk TCR/BCR CSV or TSV tables
  • inline rows for small tests

Outputs

OutputDescription
sample_summariesClonotype counts, templates, expansion counts, diversity, rarefaction summaries
v_gene_usageV gene counts and frequencies
j_gene_usageJ gene counts and frequencies
top_clonotypesTop clones and expansion classes
clone_state_linksClonotype-to-cell-state summaries
overlapCross-sample Jaccard and shared clonotype fractions
public_privatePublic/private clone classification
cdr3_length_distributionCDR3 length frequencies by sample and chain
amino_acid_summaryMean CDR3 length and amino-acid property summaries
positional_entropyHighest-entropy CDR3 positions with top amino acids
kmer_usageTop CDR3 k-mer motifs by sample and chain
vj_pairingV/J pairing frequencies
sample_distancePairwise repertoire Jaccard distances
sample_embeddingDependency-free sample distance embedding
barcode_clone_mapBarcode-level clone joins with sample, tissue, timepoint, group, chains, and ambiguity flags
clone_occupancyClone occupancy across barcoded cells, samples, states, and chains
clonal_biasDominant clone associations across state, sample, tissue, timepoint, and group
clone_alluvialClone-flow edges across sample, tissue, timepoint, and cell state
clone_networkClonotype-to-cell-state network nodes and edges
doublet_flagsAmbiguous VDJ barcodes with multiple clonotypes, excess chains, or non-productive contigs

Diversity metrics

  • Shannon entropy
  • inverse Simpson index
  • clonal evenness
  • subsampled rarefaction over observed clonotypes

Advanced sequence summaries

The advanced repertoire slice adds CDR3 sequence summaries used in immunarch/scRepertoire-style review: length distributions, amino-acid property summaries, positional entropy, k-mer usage, V/J pairing, and repertoire distance embeddings. These are descriptive research outputs and do not identify antigen specificity by themselves.

Clone-aware single-cell summaries

For single-cell VDJ tables with barcodes, the pipeline preserves barcode-level joins and reports clone occupancy, clonal bias, clone-flow edges, clone-state networks, and ambiguous VDJ barcode flags. These outputs support clone-aware review without wrapping scRepertoire code.

Reproducibility

Run records include clonotype definition, diversity estimators, advanced repertoire summary names, clone-aware summary names, and minimum clone size.