Adaptive Pipeline

Converts Adaptive Biotechnologies’ immunoSEQ v2 exports into standard AIRR TSVs and, in the same pass, reconstructs each clonotype’s full variable-domain amino acid sequence (sequence_aa) from its germline V/J genes and CDR3.

Web app slug: adaptive-pipeline

---

When to use

immunoSEQ exports use Adaptive’s own column layout and gene-naming convention (e.g. TCRBV07-02), not AIRR. Use this pipeline when you have raw immunoSEQ clonotype TSVs and need them as AIRR files — with a full reconstructed protein sequence — ready for downstream Pseudo-Sequence Alignment, Clustering, or search.

Note

Only TRA and TRB chains (human) are supported. The chain type is enforced — other chains are rejected.

---

How it works

immunoSEQ gives you gene calls and a CDR3 per clonotype, but not the full receptor protein. The pipeline normalizes Adaptive’s format to AIRR, then rebuilds the variable domain the same way a receptor is actually made — germline V, then the junction, then germline J:

sequence_aa = V germline (IMGT, up to position 103, gaps removed)
            + junction_aa (CDR3, conserved C … F/W)
            + J germline (from the J motif onward)

Each input TSV is processed independently through four stages:

1. Sort — clonotypes are sorted by abundance (count (templates/reads) descending), then by CDR3 length, so the most abundant clones lead.

2. Convert to AIRR — Adaptive’s v2 columns are mapped onto the AIRR schema (v_call, j_call, junction_aa, productive, …). At this step it keeps only productive, in-frame rearrangements, applies a J-motif sanity check, and resolves gene ambiguity (TRA has no D segment).

3. Clean & normalize genes — sequence_id is tidied; rows must have a non-empty V call, J call, and junction and be marked productive. Gene names are upper-cased and normalized from Adaptive naming toward IMGT germline names; calls that aren’t in the IMGT reference set are stripped of allele/ambiguity suffixes and re-checked, and rows whose V/J still don’t resolve are dropped. Junctions shorter than 5 aa or containing invalid amino acids are removed, and the nucleotide junction is recovered by locating the CDR3’s reading frame in the raw nucleotide sequence (reverse-complement aware).

4. Assemble sequence_aa — for each surviving row, the full amino-acid sequence is stitched from IMGT germline FASTA references: the gapped germline V sequence taken up to IMGT position 103 (gaps removed), then the junction (CDR3), then the germline J sequence from its conserved motif ([VFWG]G.G) onward. Rows whose V or J gene isn’t in the reference, whose junction has invalid residues, or whose assembled sequence contains a stop (*) are dropped.

Tip

This is the same reconstruction idea as TCR to Amino Acid Sequence — germline framework/CDR regions from the V and J calls, with the hypervariable CDR3 spliced in — but starting from Adaptive’s proprietary export and assembling against IMGT germline references up to position 103.

---

Submitting a job

Navigate to Dashboard → Adaptive Pipeline → Create New Job.

Required inputs

Field	Description
Job name	A label for this run
Input bundle (zip)	A zip of one or more immunoSEQ v2 export TSVs

Each input TSV must contain Adaptive’s clonotype columns — at minimum nucleotide, aminoAcid, count (templates/reads), vMaxResolved, jMaxResolved, and dMaxResolved.

Optional inputs

Field	Default	Description
Chain Type	TRB	TRA or TRB (only these are supported)
Number of CPU threads	4	Threads for assembly
Enable parallel assembly	true	Run the sequence_aa assembly across multiple processes
Trigger Alignment DAG	false	Automatically run Pseudo-Sequence Alignment on the output

---

Output files

File	Description
<name>-airr.tsv	One AIRR TSV per input file, with normalized v_call/j_call, recovered junction, and the reconstructed sequence_aa
logs/error.txt	Per-file processing status and any failures

All outputs are packaged into a single results.zip under the run’s output prefix. When Trigger Alignment DAG is enabled, that zip is handed directly to the Pseudo-Sequence Alignment pipeline.

---

Troubleshooting

Validation fails: missing required columns The input isn’t a clonotype-level immunoSEQ v2 export. When exporting from immunoSEQ Analyzer, choose the per-rearrangement clonotype export (not a Summary table) so columns like nucleotide, aminoAcid, and vMaxResolved are present.

Many rows dropped / empty sequence_aa Reconstruction is skipped when a row isn’t productive, its junction is too short or has invalid amino acids, or its V/J gene can’t be matched to the IMGT germline reference. Non-standard or unresolved Adaptive gene names are the most common cause — check logs/error.txt for the per-file counts and reasons.

Unsupported chain Only TRA and TRB are accepted. IGH/IGK/IGL and other chains will fail the job.