TakaraBio to AIRR

Turns the output folder of Takara Bio’s Cogent NGS Immune Profiler software into AIRR files MyImmune can consume: it picks out the TRUST4 AIRR tables from the report bundle and adds the sequence_aa column they are missing.

Web app slug: takarabio-to-airr

---

When to use

Cogent NGS Immune Profiler runs TRUST4 under the hood and writes AIRR-formatted tables, but those tables carry only nucleotide sequences (sequence, sequence_alignment) and the amino acid junction (junction_aa) — there is no full-length amino acid sequence column. Nearly every downstream MyImmune pipeline (Pseudo-Sequence Alignment, Clustering, Sequence Search) expects sequence_aa.

Use this pipeline as the first step after a Cogent run: hand it the whole output folder as a zip and it returns just the relevant AIRR files, each with sequence_aa filled in.

Note

This is a format-completion step, not a re-analysis. Every original column and value is preserved as-is — the pipeline only appends one column.

---

How it works

The full amino acid sequence isn’t stored, but it can be recovered: the nucleotide sequence is already there, and junction_aa acts as a known landmark that tells you which reading frame is the right one.

For each row:

1. Pick the nucleotide sequence — sequence_alignment is used when present, otherwise sequence. IMGT gap characters (. and -) are stripped out.

2. Translate in all three forward frames — using the standard genetic code. Unrecognized codons become X; stop codons become _.

3. Choose the frame that contains junction_aa — the CDR3 junction is known from the input, so the correct reading frame is the one whose translation contains it as a substring. That translation is written to sequence_aa.

sequence_aa = translate(sequence_alignment or sequence)
              in the forward frame whose product contains junction_aa

Caution

Rows where no frame’s translation contains junction_aa, or where the nucleotide sequence or junction_aa is empty, are dropped rather than emitted with a guessed frame. They are counted in the run’s summary.tsv (see Output files).

Which files get converted

The pipeline walks the extracted folder recursively and picks up every .csv / .tsv whose header contains all three of sequence, sequence_alignment, and junction_aa. Everything else in the Cogent output bundle — reports, QC tables, summaries, FASTAs — is ignored and does not appear in the results.

Each converted file keeps its original delimiter and its path relative to the common parent of the matched files, so the per-sample folder structure of the Cogent output is preserved in the output zip.

---

Submitting a job

Navigate to Dashboard → TakaraBio to AIRR → Create New Job.

Required inputs

Field	Description
Input bundle (zip)	A zip of the Cogent NGS Immune Profiler output folder

Optional inputs

Field	Default	Description
Job name	auto-generated	A label for this run

Tip

This pipeline has no analysis parameters to tune, so it has no presets — Create New Job takes you straight to the upload form. You can zip the Cogent output folder as-is; there is no need to hunt for the AIRR tables inside it.

---

Output files

All outputs are packaged into a single results.zip under the run’s output prefix.

File	Description
<original/relative/path>.tsv	One converted file per matched input AIRR table, with sequence_aa appended and all original columns preserved
summary.tsv	One row per converted file

summary.tsv reports how much of each file survived conversion:

Column	Meaning
filename	Path of the file, relative to the output root
total_row_count	Rows read from the input file
errors	Rows dropped because sequence_aa could not be recovered

---

Troubleshooting

Job fails with “No .csv or .tsv file with the required columns found” Nothing in the uploaded zip had all of sequence, sequence_alignment, and junction_aa in its header. Confirm you zipped the Cogent NGS Immune Profiler output folder (the one holding the TRUST4 AIRR tables) and that the zip isn’t nested inside another wrapper folder that was excluded during upload.

A sample is missing from the results Its table didn’t have all three required columns, so it was skipped — silently, and without failing the job. Run head -1 file.tsv on the input to check the header. To catch this before running, use the input bundle preflight, which flags exactly these files at upload time.

High errors count in summary.tsv Those rows had an empty nucleotide sequence or junction_aa, or no forward reading frame whose translation contained junction_aa. The latter usually points at rows where junction_aa was called against a different sequence than the one stored in the row — non-productive or partially assembled contigs are the common case.