YosefLab / velovi

https://velovi.readthedocs.io/
BSD 3-Clause "New" or "Revised" License
36 stars 7 forks source link

Reconstruction likelihood #30

Open ttgump opened 8 months ago

ttgump commented 8 months ago

Hi, In other scVI models, it uses negative binomial (NB) or zero-inflated negative binomial (ZINB) to characterize discrete count of mRNA data directly. So I just want to know why you choose normal distribution to model unspliced and spliced mRNA counts in veloVI model but not NB or ZINB?