Issue with betabinomial estimation?

[danieljrichard]

Hello,
I've been using your software to much effect for scRNA for sometime now, and was hoping to bridge into scATAC. However, I'm running into an issue with the beta binomial estimation step. Namely, after sampling 500K sites (or even more) from the basecell outputs, I get the following error - which appears to becoming from the VGAM call:

Error in checkwz(wz, M = M, trace = trace, wzepsilon = control$wzepsilon) :
Some elements in the working weights variable 'wz' are not finite

Digging into it, the alt_CC and ref_CC vectors and manually calling the beta_binom_params function, it seems to work with some subsets but not others of the vector.
When I force all alt_CC values greater than 1 to 1, however, it runs and gives me parameters. Same with alt_BC and ref_BC. However, and perhaps not-surprisingly, when I run the rest of the pipeline with these estimated parameters, the majority of my mutations have a 'BetaBin_problem' flag after step2.

Any advice? I'm working with mouse, so I don't think using the default SComatic model parameters would work here.

[ArthurDondi]

Hi,
Did you by any chance find out how to fix this problem?
Thanks!