High-resolution proteomics and machine-learning identify protein classifiers of honey made by Sicilian black honeybees (Apis mellifera ssp. sicula)

Abstract

Apis mellifera ssp. sicula, also known as the Sicilian black honeybee, is a Slow Food Presidium that produces honey with outstanding nutraceutical properties, including high antioxidant capacity. In this study, we used high-resolution proteomics to profile the honey produced by sicula and identify protein classifiers that distinguish it from that made by the more common Italian honeybee (Apis mellifera ssp. ligustica). We profiled the honey proteome of genetically pure sicula and ligustica honeybees bred in the same geographical area, so that chemical differences in their honey only reflected the genetic background of the two subspecies, rather than botanical environment. Differentially abundant proteins were validated in sicula and ligustica honeys of different origin, by using the so-called "rectangular strategy", a proteomic approach commonly used for biomarker discovery in clinical proteomics. Then, machine learning was employed to identify which proteins were the most effective in distinguishing sicula and ligustica honeys. This strategy enabled the identification of two proteins, laccase-5 and venome serine protease 34 isoform X2, that were fully effective in predicting whether honey was made by sicula or ligustica honeybees. In conclusion, we profiled the proteome of sicula honey, identified two protein classifiers of sicula honey in respect to ligustica, and proved that the rectangular strategy can be applied to uncover biomarkers to ascertain food authenticity.