Barcode collision in multi-sample single-cell runs

[zczali4403]

Description:

Hi, thanks for developing IsoQuant — it’s very helpful for long-read single-cell analysis.

I have a question about handling multiple samples in single-cell mode.

When running multiple independent samples together, cell barcodes may overlap across samples (e.g., the same 10x barcode appearing in different experiments). In this case, it seems IsoQuant:

treats identical barcodes from different samples as the same cell
performs UMI deduplication across samples
merges counts, losing sample identity

Questions:

Is this the expected behavior (i.e., assuming a single pooled experiment)?
Is there a recommended way to preserve sample identity in multi-sample runs?

Currently, it seems necessary to either run samples separately or manually add prefixes to barcodes.

Would it be possible to support something like a sample prefix or per-input-file grouping to avoid barcode collisions?

Thanks!

[andrewprzh]

Dear @zczali4403

Thanks for the positive feedback!

Yes, currently IsoQuant treats multiple FASTQs/BAMs as files from the same samples, which, in fact, is quite common use case (i.e. multiple runs of the same experiment).
There was a similar discussion recently #384

I will think about how this can be handled.

Best
Andrey