Honey Groß

New Pa­per Out

Au­then­ti­ca­ti­on of the bo­ta­ni­cal ori­gin of mo­no­flo­ral ho­ney by dielec­tric bar­ri­er di­schar­ge io­niza­ti­on high re­so­lu­ti­on mass spec­tro­me­try (DBDI-HRMS). Brea­ching the 6 s bar­ri­er of ana­ly­sis time

In: Food Con­trol June 2024, 160, 110330
Con­tri­bu­tors: An­drea Mas­s­aro, Car­me­la Za­co­metti, Mar­co Bra­go­lu­si, Jan Buček, Ro­ber­to Piro, Ales­san­dra Tata


  • A non-tar­ge­ted me­thod for au­then­ti­ca­ti­on of ho­ney by DBDI-HRMS.
  • No sam­ple pre­pa­ra­ti­on is re­qui­red and 6 s of time ana­ly­sis.
  • Va­li­da­ti­on of the me­thod with in­de­pen­dent batch was car­ri­ed out.
  • A dif­fe­rent in­ex­pe­ri­en­ced ope­ra­tor va­li­da­ted the me­thod.
  • High va­lues of per­for­mance in­di­ca­tors were ob­tai­ned.


Ra­pid vo­la­ti­le pro­fil­ing of mo­no­flo­ral ho­ney with dielec­tric bar­ri­er di­schar­ge io­niza­ti­on high re­so­lu­ti­on mass spec­tro­me­try (DBDI-HRMS) en­ab­led the au­then­ti­ca­ti­on of the stu­di­ed ho­neys’ flo­ral sources in as litt­le as 6 s in a sol­vent­less man­ner. The over­ar­ching goal of this stu­dy was the set-up of a ra­pid and cost ef­fec­ti­ve tool for the de­ter­mi­na­ti­on of the bo­ta­ni­cal ori­gin of mo­no­flo­ral ho­ney that can be ea­si­ly ac­ces­sed i) by bee­kee­pers to au­then­ti­ca­te and, thus, enhan­ce the va­lue of their own ho­ney and ii) by the food in­dus­try for qua­li­ty checks. To this aim, the vo­la­ti­le com­pounds of two in­de­pen­dent bat­ches of ho­neys with se­ven dif­fe­rent bo­ta­ni­cal ori­g­ins (aca­cia, dan­de­l­ion, chest­nut, rho­do­den­dron, ci­trus, sun­flower, and lin­den) were cap­tu­red by DBD-HRMS. A to­tal of 112 mo­no­flo­ral ho­neys were ana­ly­zed by th­ree dif­fe­rent ope­ra­tors. Using the spec­tral data of the first batch of ho­neys, we built up and com­pared the per­for­man­ces of th­ree dif­fe­rent clas­si­fi­ca­ti­on al­go­rith­ms: least ab­so­lu­te shrin­kage and sel­ec­tion ope­ra­tor (LASSO), par­ti­al least squa­res dis­cri­mi­nant ana­ly­sis (PLS-DA), and ran­dom fo­rest (RF). The per­for­man­ces of the th­ree clas­si­fiers were ve­ri­fied by re­pea­ted cross-va­li­da­ti­on, per­mu­ta­ti­on test, re­sub­sti­tu­ti­on into the trai­ning set, and then fi­nal­ly va­li­da­ted with a se­cond, in­de­pen­dent set of ho­neys by a dif­fe­rent, in­ex­pe­ri­en­ced ope­ra­tor. The out­co­mes of the tests were ex­pres­sed by the area un­der the cur­ve (AUC), Kap­pa sta­tis­tic, over­all ac­cu­ra­cy, and sen­si­ti­vi­ty and spe­ci­fi­ci­ty ra­tes. The mis­clas­si­fi­ca­ti­on ra­tes of the built clas­si­fiers were eva­lua­ted by com­pu­ting dif­fe­rent key in­di­ca­tors, and the re­pea­ta­bi­li­ty of the ana­ly­ti­cal me­thod was also ve­ri­fied by co­si­ne si­mi­la­ri­ty. The­se in­sights are re­le­vant for fu­ture ad­op­ti­on of the me­thod in rou­ti­ne work. We de­ter­mi­ned that the RF clas­si­fier was the most powerful in pre­dic­ting the flo­ral source of the ho­neys. The RF clas­si­fier pro­du­ced high per­for­mance va­lues when pre­dic­ting an in­de­pen­dent batch of samples ana­ly­zed by a dif­fe­rent, in­ex­pe­ri­en­ced ope­ra­tor (AUC 82.91 %, over­all ac­cu­ra­cy 81.25 %, Kap­pa 77.78 %, sen­si­ti­vi­ty 81.05 %, and spe­ci­fi­ci­ty 96.76 %). The mis­clas­si­fied ho­neys were tho­se cha­rac­te­ri­zed by the pre­sence of other nec­tars and or pol­lens, as pre­vious­ly poin­ted to by the sen­so­ry pa­nelists This pro­of-of-prin­ci­ple work war­rants a fu­ture lar­ge-sca­le stu­dy to va­li­da­te and chall­enge the me­thod with ho­neys from dif­fe­rent har­ve­sts and of dif­fe­rent geo­gra­phi­cal ori­g­ins.