alevin-fry for transcript-level quantifcation

[hwoodw03]

Dear Team, on alevin-fry quant you need to provide a tg-map input file as a transcript to gene map to carry out the quantification. However, are you able to obtain a transcript x cell matrix count instead of a gene x cell matrix count as output?

[njbernstein]

Bump on this issue. Would be great to see the underlying counts here.

[rob-p]

If you provide a transcript-to-transcript map to quantification rather than a transcript to gene map, then alevin-fry will produce a transcript-level count matrix.

However, if one does this, they should be aware of the relevant caveats. Specifically, since most short-read high-throughout single-cell RNA-seq protocols are tag-based, and exhibit a strong 3' bias, there will be a large amount of ambiguity in the UMI to transcript assignment. Depending on the resolution method, ambiguous UMIs may be discarded or (if using an EM algorithm based resolution) probabilistically assigned among multi mapping transcripts.