Stability of proteomic signals and choice reference library
Stability of proteomic signals between samples and regions was analyzed comparing the intra-specific variances of Euclidean distances between all specimens within a sample, within a region from different samples and seasons as well as between all specimens coming from different regions. In a next step we tested how reliable species identification can be done using inter-regional libraries: species RF models excluding all specimens from a certain region were used to assign these specimens to a RF model species class, respectively. A post hoc test for false positive discovery of proteomic profile-based RF models (Rossel & Martinez Arbizu, 2018a, https://github.com/pmartinezarbizu/RFtools) was applied, using a significance of <0.01 to reject a classification as potential false positive. Heatmaps presenting the distance between single specimens from different regions were created for congener pairsA. tonsa (three regions) and A. longiremi s (six regions),C. typicus (two regions) and C. hamatus (five regions),T. stylifera (one region) and T. longicornis (seven regions) as well as C. hyperboreus (three regions) and C. finmarchicus (six regions).