Fig. 1.
A, Overview of the PolySTest web service and the used statistical tests. The FDRs from the four statistical tests are combined into a unified PolySTest FDR. B, Combinations of missing values (zeroes) and present values (ones) in the case of three replicates and two conditions A and B. Given a probability of pNA = 1/2 for a missing value, each combination is found with probability 1/64. The numbers on the right count the absolute of the difference in missing values. This leads to p values of 5/16, 15/32, 3/16 and 1/32 for finding a difference of missing values of 0,1,2 and 3, respectively. C, Scheme of Miss test on an example of two conditions with three replicates where two of the values of condition A are absent. The distribution of all quantifications in the dataset defines the quantiles used as thresholds for iteratively removing values from the dataset. The differences in the number of missing values are then calculated for each threshold. Note that a different threshold also implies changing pNA. The figure also shows exemplary p values for the different quantiles. Here, Miss test would report the p value marked in bold.