Simulations for Umbrella and JT test. Time series were generated containing no trend, an upward trend or a peak/low point at time point 11. A-D show these time series without noise and E-H for jittered data. On these time series the Wilcoxon, JT and Umbrella tests were performed and p-values calculated which are shown in the figure legends. Only the Umbrella test is able to identify peaked trends in the data as they are expected from the experimental design of the metabolite data.