Fig. 3: Results of Simulation 1.

a. Simulation results under Simulation 1 null case with the four methods. The Q-Q plots of the observed \(-{\log }_{10}p\) against the expected \(-{\log }_{10}p\) for different methods are displayed across various dispersion parameters and scenarios, where p represents the p-value from the score test. False positive rates (type I error rates) for each method are displayed in the top-left corner of each panel, with colors corresponding to the legend. The light gray region represents the 95% error band, illustrating the range in which 95% of points are expected to fall under the null hypothesis, assuming the p-values follow the expected uniform distribution. b Simulation results under Simulation 1 alternative case 1, which involves both SVGs and ctSVGs. The plots display power values across a range of false discovery rates for different methods, considering various dispersion parameters and cell-type composition scenarios. c Simulation results under Simulation 1 alternative case 2, which involves only ctSVGs. The plots display power values across a range of false discovery rates for different methods, considering various dispersion parameters and cell-type composition scenarios. d Simulation results under Simulation 1 alternative case 3, which involves only SVGs. The plots display power values across a range of false discovery rates for different methods, considering various dispersion parameters. Source data are provided as a Source Data file.