Repertoire analysis TCRshows

Drop your TCR sequencing output straight in. Four common formats are detected automatically — 10x Genomics, AIRR Rearrangement, Adaptive immunoSEQ and MiXCR — and any other table can have its columns mapped by hand. One pass gives you clonality, diversity, clone-size strata, CDR3 length distribution, V/J gene usage, and the overlap and clone tracking between samples. Nothing is uploaded: parsing and computation happen in your own browser and disappear when you close the page.

1 · Import data

Drop several files at once — each is treated as one sample. Overlap and clone tracking need the pre and post samples of the same experiment dropped in together.

Drop files here, or click to choose 10x filtered_contig_annotations.csv · AIRR *.tsv · immunoSEQ *.tsv · MiXCR clones.txt · generic CSV / TSV

Clonotype definition

Chain

Rows in the clonotype table

Keep productive sequences only Drop clonotypes seen only once Rarefy to the smallest sample depth before comparing

2 · Sample metrics

How to read these

Clonality = 1 − Shannon evenness. Zero means every clonotype is equally abundant; the closer to 1, the more a few clones dominate. A rise in peripheral-blood TRB clonality after a TCR-T infusion usually means the product is expanding in vivo.

Shannon entropy H = −Σ p·ln p, which reflects both richness and evenness; evenness (Pielou) = H / ln S divides the richness out.

Simpson (1−Σp²) is the probability that two sequences drawn at random come from different clonotypes; inverse Simpson (1/Σp²) is the equivalent “effective number of clonotypes”.

D50 = how many clonotypes it takes to reach 50% cumulative frequency; smaller means more oligoclonal. The bracket gives that as a percentage of all clonotypes.

Gini, borrowed from income inequality: 0 is perfectly even, 1 perfectly concentrated. More sensitive to the top clones than Shannon.

Chao1 estimates what sequencing missed from the number of clonotypes seen once and twice, giving a lower bound on richness. Only meaningful on bulk data with counts.

3 · Clonotype distribution

Sample

Clone-size strata

Clonotypes binned by their frequency within the sample, showing how large the clones holding the repertoire are.

Cumulative frequency curve

Clonotypes ranked by abundance on the x axis (log), cumulative share on the y axis. The steeper the curve, the more oligoclonal.

CDR3 length distribution

Counted per clonotype, not weighted by templates. An unusual shift in length usually points to a library-prep or filtering problem.

Most abundant clonotypes

4 · V / J gene usage

Counted as a share of clonotypes — each clonotype once, not weighted by templates — with gene names normalised to IMGT nomenclature.

V gene usage (top 20)

J gene usage (top 20)

V-J pairing

Shade is that V-J combination’s share of all clonotypes; only the most-used V and J genes are shown.

5 · Comparing samples

Overlap metric

Overlap matrix

Clone frequency tracking

Frequency scatter (log-log)

Clonotypes above the diagonal expanded in the sample on the y axis. Those found in only one sample are drawn on the edge axes.

Shared clonotypes that changed most

Formats, definitions and limits

Supported input

Format	Typical file	Detected by	Abundance
10x Genomics V(D)J	filtered_contig_annotations.csv	barcode + chain + cdr3	Cells (deduplicated by barcode)
AIRR Rearrangement	*.tsv	v_call + junction_aa	duplicate_count / consensus_count, defaulting to 1
Adaptive immunoSEQ	*.tsv	amino_acid + templates, etc.	templates, falling back to reads
MiXCR	clones.txt / clones.tsv	aaSeqCDR3 + cloneCount	cloneCount / readCount
Generic CSV / TSV	Any	Guessed from column names; editable	The nominated count column, defaulting to 1

Definitions. Clonotypes are collapsed on V + CDR3 amino acid + J by default — collapsing on the CDR3 amino acid alone would merge identical CDR3s carried by different V genes, while collapsing on nucleotides would split synonymous variants apart. Gene names are normalised to IMGT (immunoSEQ’s TCRBV06-01 becomes TRBV6-1, and the *01 allele suffix is always dropped). Single-cell data is deduplicated by barcode, so one cell counts once per clonotype. In α/β paired mode only cells with exactly one TRA and one TRB are used; dual-α cells and cells with a single chain are reported separately in the pairing summary rather than counted as paired clonotypes.

Check sequencing depth before comparing samples. Richness metrics — clonotype count, Chao1, D50 — depend strongly on depth, and comparing them across samples that differ several-fold is meaningless. Ticking “rarefy to the smallest sample depth” randomly downsamples every sample to the same template count first, which is the comparable basis (rarefaction uses a fixed seed, so results reproduce). Clonality and Shannon evenness are less sensitive to depth, but not immune to it.
This tool only describes the repertoire statistically; it makes no inference about the antigen specificity of any clonotype. To check whether a CDR3 matches a known target, see TCRshowsdb.