Fig. 3.

48 h Synthetic Data Method Comparison—Noise Level 0.1. Comparison of TimeCycle, JTK_CYCLE, RAIN, SW1PerS and GeneCycle results for synthetic data of various waveform shapes. LOWER TRIANGLE: scatterplot of $-{\hspace{0.17em}\text{log}\hspace{0.17em}}_{10}P$ for each synthetic gene. The larger colored points represent the average $-{\hspace{0.17em}\text{log}\hspace{0.17em}}_{10}P$ designated by waveform shape. DIAGONAL: Histogram of $-{\hspace{0.17em}\text{log}\hspace{0.17em}}_{10}P$ for each synthetic gene by method. UPPER TRIANGLE: Classification accuracy comparison by waveform category. Darker bars correspond to the method listed in the row; lighter bars correspond to the method listed in the column. Each value corresponds to the fraction of correctly classified genes (true positive for cycling/non-cycling as appropriate) for a given waveform type, using a classification threshold obtained from Youden’s J index for the ROC across all genes. * SW1PerS does not compute a P-value, but rather a periodicity score. $†$ GeneCycle results of P = 0 were set to machine precision ($2.2\times {10}^{-16}$) for visualization purposes.