Fig. 1
From: baredSC: Bayesian approach to retrieve expression distribution of single-cell data
Poisson is a good approximation to explain variance in droplet-based single-cell RNA-seq. In each dataset, control dataset (A) or real single-cell experiment (B), the normalized variance estimator is plotted in function of the estimated mean expression. Each dot is a gene. A Gaussian kernel smoother was applied to evaluate the tendency of the data (red line) and the error around the Gaussian kernel smoothing was estimated (yellow area). The expected variance in the Poisson approximation is a straight line (blue) whereas the expected variance in the negative binomial approximation is a quadratic curve (green). Both axes are in log scale