Fig. 3

Gene-specific analysis, including model selection and likelihood ratio tests, across datasets. A The proportion of genes that prefer each of the four count models based on model selection with AIC (y-axis) is displayed for each dataset (x-axis). The four models include Poisson (Aquamarine), negative binomial (NB; Tangerine), zero-inflated Poisson (ZIP; Polo Blue), and zero-inflated negative binomial (ZINB; pink). The Poisson model and the negative binomial model are the preferred models across datasets. B The proportion of genes that are significant in each of the four likelihood ratio tests (y-axis) is displayed for each dataset (x-axis). The four LRT tests include the test on P vs ZIP (salmon), P vs NB (green), NB vs ZINB (cyan), and ZIP vs ZINB (purple)