Figure 1

Scheme of the procedure to calculate sum of ranking differences. The input matrix contains similarity measures (n = 8) in the columns and molecules (m = 99) in the rows. A reference column (golden standard, here: average of the eight similarity measures) is added in the data fusion step (red). Then, all columns are doubled (green) and the molecules in each column are ranked by increasing magnitude (columns r1, r2, … rn). The differences (yellow columns) are calculated for each similarity measure and each molecule (i.e. each cell) between its rank (r1₁, r1₂ to rn_m) and the rank assigned by the known reference method (rR = q₁, q₂, … q_m). In the last step, the absolute values of the differences are summed up for each measure to give the final SRD values, which are to be compared. The smaller SRD means proximity to the reference, the smaller the better.