Fig. 4

Inference of subclone evolutionary dynamics. Computational modeling predicts the relationship between clone intermixing and clone size, and the time of clone appearance and selective advantage. Clone mixing level increases to a transient high at intermediate times as the subclone expands, and decreases as the clone takes over the tumor or is lost (time increases along the columns). The strength of selection experienced by the subclone determines the speed and maximal extent of subclone mixing (selection increases down the rows). Black dots are empirically measured mixing values in primary tumors, and are plotted for the parameter values that explain the data with maximum posterior probability (see Methods). The inner panels (dashed circles) are representative simulation snapshots indicating the expected clonal intermixing. Each panel shows results from 1000 independent realizations; relative growth ratios of the mutant to wild type are 1, 1.75, 3.0, 8.0 for neutral, weak, intermediate, and strong selection, respectively; the initial mutant frequency is 0.01 and the final tumor size is 10⁴