Classification success

Finally, we tested the identification success based on MALDI-TOF MS data for each specimen in the data set by excluding the respective specimen and using the remaining reference data to identify it.

Overall, 93% of the specimens (n=775) were identified correctly and 86% (n=721) were accepted as correctly classified by thepost-hoc test (Fig. 3A). Identification for specimens of the classes Ascidacea, Teleostei, Elasmobranchii, Echinoidea, Ophiuroidea, Asteroidea, Bivalvia and Gastropoda resulted in success rates of more than 90%. For classes Cephalopoda and Thecostraca the identification success was still above 85%. Success rates lower than 80% were not recorded. Of the 61 misclassified specimens, 15 were assigned to the false species and recorded as correct identifications by thepost-hoc test. Of all misclassified specimens, two were assigned to congeneric classes and rated as true positives by the post-hoctest, meaning these would have been misclassified and remain unrecognized.