Extending Bayesian Modelling of RNA Velocity

This paper introduces an alternative method for modeling RNA velocity within the Bayesian framework, employing zero-inflated distributions without the need for artificial preprocessing to handle RNA counts. Through a comparative analysis conducted on a real dataset, we illustrate the performance of our approach, showcasing outcomes comparable to those achieved with assumptions of Negative Binomial data on preprocessed observations. Our proposed model eliminates the requirement for arbitrary data filtering, thereby demonstrating its effectiveness in capturing the underlying biological dynamics.