Fig. 4. SIMILE identifies the largest number of similar pairs and 88% and 89% of the pairs of spectra identified by SIMILE in positive and negative ionization mode were not found by any other algorithm.

Pairs of spectra identified by each algorithm for positive (a–c) and negative (d–f) ionization modes can be seen in these UpSet plots. The total number of pairs identified by each algorithm is shown in the bar chart to the left (a and d), and the unique pairs identified by each algorithm along with their corresponding structural similarity are shown in (b, c) and (e, f). Boxen plots (b, e) show several quantiles of the maximum common substructure (MCS) for each algorithm to approximate a distribution³⁸.