Species identification of shrimps, lobsters and crayfish using peptide-based liquid chromatography low-resolution mass spectrometry and random forest model

Abstract Seafood, including crustaceans, provides a high-quality animal-origin protein source with globally increasing consumption rates. However, international seafood trade is frequently confronted with issues like commercial fraud and safety risks for human health, partly due to incorrect species identification of aquatic food products. To distinguish between crustacean species through the identification of species-specific peptides, an untargeted liquid chromatography low-resolution mass spectrometry (LC-LRMS) method was developed and validated. Programmed algorithms were applied to identify marker candidates from peptide profiles. A targeted multiple reaction monitoring (MRM) method was established to verify the suitability of selected candidates as crustacean biomarkers related to the used dataset. An additional random forest model was built to determine unknown crustacean species based on an LC-LRMS raw data training set. In total, 49 out of 150 selected peptides were identified as species-specific biomarkers based on the monitored dataset and those are able to differentiate 14 crustacean species. To further evaluate specificity, additional commercial samples from several crustacean, mussel, insect and fish species were tested. De novo sequencing of LC-LRMS and comparing high-resolution mass spectrometry (HRMS) data mostly showed similar results concerning the proposed amino acid sequence and average local confidence score. Selected peptide markers were synthesized and experimentally confirmed. The random forest model exemplarily demonstrated the correct identification in an unseen test dataset based on plurality vote. The results show the feasibility of solving authenticity questions, e.g. to identify unknown crustacean species, by applying alternative procedures without using HRMS instruments, requiring a higher effort. At the same time, the results prove that certain limitations cannot be overcome using LRMS devices.