Question on base-calling implementation (A/T/G/C) in PerturbView ISS pipeline

[chenzhaolab2023]

Hi @taka910 and other PerturbView authors,

Thank you for creating this wonderful tool! We are trying to understand the base-calling/decoding step in more detail. Specifically, we are looking for where (and how) the pipeline assigns detected sequencing signals to A / T / G / C at each cycle. In particular:

1. Which part of the code performs base calling from multi-channel fluorescence images?
2. Is base identity determined via per-cycle channel intensity comparison (e.g., argmax across A/T/G/C channels), or using a different scoring / normalization approach?
3. How are ambiguous or low-confidence base calls handled (e.g., thresholds, masking, N calls)?
4. Is this logic implemented directly in the public repository, or assumed to be performed upstream as a preprocessing step before running PerturbView?

We may have missed it, but we were unable to identify a clearly defined base-calling or ISS decoding module in the repo. Pointers to the relevant scripts, functions, or documentation would be very helpful.

Thanks for making this work publicly available.
Chen