Triangular Test of Amanita Mushrooms by Using Electronic Nose and Sensory Panel

foods

Article Triangular Test of Amanita Mushrooms by Using Electronic Nose and Sensory Panel

Francisco Portalo-Calero 1 , Patricia Arroyo 1 , José Ignacio Suárez 1 and Jesús Lozano 1,2,*

1 Escuela de Ingenierías Industriales, Universidad de Extremadura, 06006 Badajoz, Spain; [email protected] (F.P.-C.); [email protected] (P.A.); [email protected] (J.I.S.) 2 Instituto Universitario de Investigación en Recursos Agrarios (INURA), Universidad de Extremadura, Avd. De la Investigación, 06006 Badajoz, Spain * Correspondence: [email protected]; Tel.: +34-924-289-300

Received: 19 August 2019; Accepted: 10 September 2019; Published: 14 September 2019

Abstract: This work aims to advance understanding of the differentiation of mushroom species through electronic devices that use sensors of various technologies and techniques for pattern recognition, comparing mainly volatile substances that emanate from them. In this first phase, the capacity of human olfaction to differentiate between the smell released by different wild mushrooms of the genus Amanita was analyzed by means of a triangular sensory test, comparing later the data to those obtained for the same samples with an electronic nose in a similar test. The results, still very preliminary, encourage imagining the wide application that these techniques will have and the feedback that this application can suppose for the training of the sense of human olfaction.

Keywords: smell; sensory analysis; triangular test; mycology; Amanita; electronic nose

1. Introduction Recent studies have seemed to indicate that the fact that humans have fewer olfactory receptors than some animals [1] does not necessarily mean that they have a worse olfactory sense [2]. It is currently considered that human olfaction is capable of detecting many more of the 10,000 smells than popular wisdom suggests [3]. This contradicts, to a certain extent, its long-suffering reputation as being only the fifth sense [4], since our ancestors abandoned the condition of hunter-gatherers. In the 19th century, the French scientist Paul Broca considered the human olfactory system to be primitive and its evolution to be inversely proportional to that of intelligence [2].There is also an added cultural problem, and according to References [4,5], this indicates that “...One could speculate that we have lost the ability to describe odors with the shift to agriculture...the inability to name odors is not a biological limitation, but also a cultural one”. Possibly this limitation in defining odors through language has favored generalization when classifying them. For example, sensory properties as broad as sweet, fruity, syrupy, earthy, and antiseptic have long been used to define organoleptic characteristics in the classification of mushrooms [6]. On the other hand, it is known that certain animals develop the capacity to detect volatile substances and thus find traces of people, smell diseases [7], locate explosives [8], etc. Among the natural aromas that they are able to differentiate between are those released by some types of mushrooms, which are used traditionally in their harvest once trained for it. This circumstance suggests the possible existence of different odorous nuances between the different species of mushrooms beyond those that are currently recognized by the human sense of smell. The literature on fungi, mainly the field guides that include a card for each species evaluated, contains a macroscopic description of mushrooms, where the smell attribute is generally found within the meat section. However, although they sometimes provide important data for the classification of certain species, the description is usually very vague and generalist, so it cannot be considered, by itself,

Foods 2019, 8, 414; doi:10.3390/foods8090414 www.mdpi.com/journal/foods Foods 2019,, 8,, 414414 22 of of 19 20 classification of certain species, the description is usually very vague and generalist, so it cannot be definitiveconsidered, in by determining itself, definitive the species in determining in question. the If species we focus in onquestion. a review If ofwe the focus information on a review provided of the byinformation these guides provided on the by wild these mushrooms guides on analyzed the wild in mushrooms this work, in analyzed all of the in genus this work, Amanita, in all which of the is locatedgenus Amanita, in the southwest which is of located the Iberian in the Peninsula, southwest uniformity of the Iberian has been Peninsula, detected uniformity in odor assessments. has been However,detected in this odor is assessments. not complete, Howeve and ther, this lexical is not deficiency complete, is and noted the in le thexical definition deficiency of is the noted olfactory in the characterdefinition ofof eachthe olfactory of the species character of mushrooms. of each of the For species example, of mushrooms. the Amanita For caesarea, example, a the well-known Amanita andcaesarea, gastronomically a well-known appreciated and gastronomically mushroom, apprecia is indicatedted mushroom, in different is guides indicated as follows: in different “little guides that isas appreciable”follows: “little [9 ],that “pleasant is appreciable” mushroom [9], smell”“pleasant [10], mushroom “smooth and smell” pleasant [10], odor”“smooth [11 and], or pleasant “almost nil”odor” [12 [11],]. or “almost nil” [12]. The The use use of of this this methodology methodologyof of descriptors,descriptors, soso generalgeneral andand wellwell known,known, can be eefficientfficient among those speciesspecies inin which which there there is ais notablea notable odorous odorous diff erence.difference. Nevertheless, Nevertheless, they greatlythey greatly limit thelimit ability the toability distinguish to distinguish one species one fromspecies another from whenanother the when character the character of its odor of is its nearby. odor is This nearby. need forThis taxonomic need for classification,taxonomic classification, apart from the apart scientific from point the ofscientific view, is absolutelypoint of view, necessary is absolutely for a hypothetical necessary consumer, for a ashypothetical there are edibleconsumer, individuals as there and are othersedible withindividu a highals degreeand others of toxicity. with a Obviously,high degree mycologists, of toxicity. amateurs,Obviously, andmycologists, experts use amateurs, other markers and experts in their use classification, other markers but sometimesin their classification, the difficulty but of unequivocalsometimes the identification difficulty of aunequivocal fungus through identification its phenotypic of a characteristicsfungus through (morphology its phenotypic and physiology)characteristics is extremely(morphology diffi cultand [physiology)13]. However, is forextr example,emely difficult in recent [13]. years However, in Spain, for approximately example, in 0.2%recent of years consultations in Spain, inapproximately which the Toxicological 0.2% of consultations Information in Service which issuedthe Toxicological a report [14 Information] have been relatedService toissued mushrooms. a report Although [14] have this been did related not always to mushrooms. mean serious Although poisoning, this in did some not cases always (Figure mean1), theserious consequences poisoning, were in some dramatic cases [(Figure15]. 1), the consequences were dramatic [15].

Figure 1. Amanita phalloides ,, cause cause of of the deadly phalloidin syndrome.

At this point, two questions could be asked: Could humans really differentiate certain species of At this point, two questions could be asked: Could humans really differentiate certain species of mushrooms by smell? In addition, could artificial systems be useful in this task? Sensory evaluation was mushrooms by smell? In addition, could artificial systems be useful in this task? Sensory evaluation used to try to answer the first question. The use of this standardized tool, which is of great importance in was used to try to answer the first question. The use of this standardized tool, which is of great food research and development [16] and which has already been used in the recognition of mushroom importance in food research and development [16] and which has already been used in the characteristics [17], has proven to be a good choice in obtaining the indicators of human capacity to recognition of mushroom characteristics [17], has proven to be a good choice in obtaining the diindicatorsfferentiate of between human odors.capacity Recent to differentiate studies have betw usedeen triangular odors. Recent testing, studie alongs withhave otherused techniques,triangular totesting, identify along key with aromatic other compounds techniques, as to well identify as ketones key aromatic and alcohols compounds in some typesas well of fungias ketones [18].With and respectalcohols to in thesome second types questionof fungi [18]. raised, With we respect can point to the out second that Referencequestion raised, [19] demonstrated we can point thatout that the Penicillium speciesReference of [19] fungi demonstrated of the genus that the speciescould of fungi be discriminated of the genus Penicillium between by could means be ofdiscriminated an analysis between by means of an analysis of volatile secondary metabolites. Moreover, a gas

Foods 2019, 8, 414 3 of 20 of volatile secondary metabolites. Moreover, a gas chromatography–mass spectrometry (GC–MS) evaluation of different species of fungi [20–22] made evident that the chemical structures of many of these metabolites are similar to those of known fragrance chemicals. In Reference [6], a table showed more than 150 types of aromatic components present in mushrooms related to one or several odors. On the other hand, the field of research on artificial sensory systems, more specifically that developed by so-called electronic noses, has demonstrated its potential to identify and solve problems in cosmetic production processes, food industries [23], chemical engineering, environmental monitoring [24], explosives detection, and medical diagnostics and bioprocesses [25]. There has also been electronic nose analysis in the literature that has differentiated between the hedonic tones of mushrooms [26], and there have been electronic noses in conjunction with GC–MS techniques to determine the degree of drying of cultivated mushrooms [27]. It is important to note that the applications developed with these devices are increasingly accessible and portable [28,29]. This work is a first step within a research project that aims to advance understanding of the differentiation of mushroom species through electronic equipment and artificial intelligence (AI) techniques, comparing mainly volatile substances that emanate from them. Later on, the aim is to develop devices that discriminate between wild and commercial mushrooms, taking into account the possible harmful substances present in these mushrooms. In this experiment, once the presence of characteristic components of the aroma of mushrooms was confirmed by GC–MS techniques, the same samples analyzed in the sensory test were subjected to an analysis using an electronic nose. Then, a comparative evaluation of the results was subsequently carried out.

2. Materials and Methods

2.1. Samples Before starting the collection work, the possible pathologies derived from the manipulation or inhalation of spores of toxic mushrooms were studied, as well as the possible influence that this manipulation could have on the information obtained. Once the corresponding consultations had been carried out, it was clearly established that the handling of toxic mushrooms does not, in itself, entail a risk to the health of the people who smell or touch them [30]. The use of neutral-odor latex gloves was established throughout the preparation process in order to avoid allergic reactions or possible contamination of the samples. The statistical population upon which this work was carried out was defined by 6 species of mushrooms of the genus Amanita (Table1). These species are present in di fferent areas of the southwest Iberian Peninsula (Table2) and were collected during the autumn of 2018. At that moment, all selected mushrooms were classified by species and stage of ripeness (young, optimal, and mature) by at least two independent experts. For the tests, the hats [31] from individuals of optimal mushrooms were chosen. In a period of time not greater than four hours from the harvest, they were put under a process of freezing to 24 C. Prior to freezing, they were quartered, and each sample was weighed and stored in coded − ◦ odorless bags. The major parts were allocated for triangular sensory analysis and the remaining parts for triangular analysis with an electronic nose. Some individuals were assigned to perform a GC–MS analysis of all mushroom species participating in the study. Foods 2019, 8, 414 4 of 19 FoodsFoodsFoods 20192019 2019,, 88, ,,8 414414, 414 44 4ofof of 1919 19 Table 1. Edibility/toxicity and aroma of the Amanita species analyzed. Table 1. Edibility/toxicity and aroma of the Amanita speciesspecies analyzed.analyzed. Table 1. Edibility/toxicity and aroma of the Amanita species analyzed. A. A. A. A. muscaria A. caesarea Mushroom rubescens pantherina A. citrina (C) phalloides FoodsFoods 2019A.A.Foods Foods ,2019 8, 414 2019, 20198 , 414, 8, ,8 414, 414A. A. A.A. 4 of 419 of 4 194of of 19 19 (M) (Ca) Foods 2019, 8, 414 A.A. muscaria muscaria A.A. caesarea caesarea4 of 19 (R) 1 (P) (Ph) MushroomMushroom rubescensrubescensrubescens pantherinapantherinapantherina A.A. citrina citrina (C)(C) (C) phalloidesphalloidesphalloides (M)(M)(M) (Ca)(Ca)(Ca) (R)(R) 11 1 (P)(P)Table TableTable 1. Edibility/toxicity 1. 1. Edibility/toxicity Edibility/toxicity and and aromaand aroma aroma of the of of theAmanita the Amanita Amanita(Ph)(Ph) species species species analyzed. analyzed. analyzed. (R) TableTable (P)1. Edibility/toxicity 1. Edibility/toxicity and and aroma aroma of the of Amanitathe Amanita species species analyzed.(Ph) analyzed. A. A. A.A. A. A. A.A. A. A. A.A. Image A. A. A. muscariaA. muscariaA.A. muscaria muscaria A. A. caesareaA. caesareaA.A. caesarea caesarea MushroomMushroomMushroomMushroom rubescens rubescens rubescensrubescens pantherina pantherinapantherinapantherina A. citrina A. citrinaA.A. (C)citrina citrina A. (C) muscaria (C) (C) phalloidesphalloidesphalloidesphalloides A. caesarea Mushroom rubescens pantherina A. citrina (C) (M) (M)(M) phalloides(M) (Ca) (Ca)(Ca) (Ca) 1 1 1 1 (M) (Ca) ImageImageImage (R)(R) 1 (R) (R)(R) (P) (P) (P) (P)(P) (Ph) (Ph) (Ph)(Ph) (Ph)

Edibility/to xicity ImageImage Image ImageImage

Edibility/toEdibility/to Odorless Foods 2019, 8, 414 Smell Odorless4 of 20 Soft Unpleasant Unpleasant Nice xicityxicityxicity (young) Note: Amanita rubescens requires previous cooking. Edibility/toEdibility/toEdibility/toEdibility/toEdibility/to xicityxicity xicityxicity OdorlessOdorless SmellSmell xicityOdorlessOdorlessOdorless Table 1. Soft Soft SoftEdibility / Unpleasant Unpleasanttoxicity Unpleasant and aromaUnpleasantUnpleasantUnpleasant of the Amanita species analyzed.NiceNice (young)(young)(young)

Note:Note: Amanita Amanita rubescens rubescensA. requiresrequires requires previousprevious previous cooking.cooking. cooking. OdorlessOdorlessOdorless Odorless OdorlessA. SmellSmell Smell SmellSmellA.Odorless rubescensOdorless Odorless Odorless Odorless Soft Soft Soft Soft Unpleasant Soft Unpleasant A. Unpleasant citrina Unpleasant Unpleasant Unpleasant Unpleasant A. Unpleasant muscaria UnpleasantUnpleasant Nice NiceNice Nice Nice Mushroom pantherina (young)(young) phalloides(young)(young) (young) A. caesarea (Ca) (R) 1 (C) (M) Note:Note: AmanitaNote: Amanita(P)Note:Note: Amanita rubescens Amanita rubescensAmanita rubescens requires rubescens rubescensrequires requiresprevious requires previousrequires previous cooking. previous previouscooking. cooking. cooking. cooking. (Ph)

Image

Edibility/toxicity

Odorless Table 2. Most significant vegetation type and collection areas. Smell Odorless Soft Unpleasant Unpleasant Nice (young) Collection Area Vegetation City Note: Amanita rubescens requires previous cooking. Serra de S. Mamede Cork oak, holm oak, scrub Portalegre (Portugal) Comarca de La Vera Oak and chestnut, scrub Cáceres (Spain) Sierra de Aracena Cork oak, holm oak, scrub Huelva (Spain)

Comarca de Badajoz Holm oak, eucalyptus, scrub Badajoz (Spain) Monte dos Arneiros Cork oak, scrub Montemor-o-Novo (Portugal) TableTable 2. 2.Most Most significant significant vegetation vegetation type and type collection and areas. collection areas. TableTable 2. 2. Most MostTable significant significantTable 2.Table MostTable 2. Mostvegetation 2.significant vegetation2. Most Most significant significant significant vegetationtype type vegetation and andvegetation vegetation collectiontype collection type and type typecollectionand areas. areas.and collectionand collection collection areas. areas. areas. areas. Sierra Suroeste Chestnut, holm oak, cork oak, scrub Badajoz (Spain) Collection Area Vegetation City CollectionCollection AreaCollection Collection AreaCollectionCollection Area Area Area Area Vegetation Vegetation Vegetation Vegetation Vegetation Vegetation City City City City City City CollectionCollection SerraArea Area de S. Mamede Vegetation VegetationCork oak, holm oak, scrub Portalegre City City (Portugal) For the tests, the hats [31] from individuals of optimal mushrooms were chosen. In a period of SerraSerra deSerra S.Serra de Mamede S.de de Mamede S. S. Mamede Mamede Cork Cork oak, Cork Cork oak,holm oak, oak,holm oak, holm holm oak,scrub oak, oak,scrub scrub scrub Portalegre Portalegre Portalegre Portalegre (Portugal) (Portugal) (Portugal) (Portugal) SerraSerraSerra dede de S.S. S. MamedeMamede MamedeComarca de La Vera Cork Cork Cork oak,oak, oak, Oakholmholm holm and oak,oak, oak, chestnut, scrubscrub scrub scrub Portalegre Portalegre Portalegre Cáceres (Portugal)(Portugal) (Portugal) (Spain) time not greater than four hours from the harvest, they were put under a process of freezing to −24 SerraComarca de S.Comarca MamedeComarca Comarcade La de Vera Lade de VeraLa La Vera Vera Cork Oak Oakand oak, Oak Oakchestnut,and holm and chestnut,and chestnut, oak, chestnut,scrub scrub scrub scrub scrub Cáceres Portalegre Cáceres Cáceres(Spain) Cáceres (Spain) (Spain) (Portugal)(Spain) ComarcaComarca de de La LaSierra Vera Vera de Aracena Oak Oak and and Cork chestnut, chestnut, oak, ho scrublm scrub oak, scrub Cáceres Cáceres Huelva (Spain) (Spain) (Spain) ComarcaSierra deSierra de LaSierra SierraAracena de Vera Aracenade de Aracena Aracena CorkOak Cork oak, and Cork Cork oak,ho chestnut, lmoak, oak,ho oak,lm ho ho lmoak,scrublm scruboak, oak,scrub scrub scrub Huelva HuelvaC Huelva(Spain)á Huelvaceres (Spain)°C. (Spain) (Spain) (Spain)Prior to freezing, they were quartered, and each sample was weighed and stored in coded SierraSierra dede AracenaAracenaComarca de Badajoz Cork Cork oak,oak, Holm hoholmlm oak, oak,oak, eucalyptus, scrubscrub scrub Huelva Huelva Badajoz (Spain)(Spain) (Spain) Sierra deSierra AracenaComarca deComarca AracenaComarca Comarcade Badajoz de Badajozde de Badajoz CorkBadajoz oak, HolmCorkholm Holm oak, oak, Holm Holm oak,eucalyptus, scrub holm oak, oak,eucalyptus, eucalyptus, oak,eucalyptus, scrub scrub scrub scrub scrub Huelva Badajoz (Spain) Badajoz HuelvaBadajoz (Spain) Badajoz (Spain)odorless (Spain) (Spain) bags. The major parts were allocated for triangular sensory analysis and the remaining parts Comarca de BadajozMonte dos Arneiros Holm oak, eucalyptus, Cork oak, scrub scrub Montemor-o-Novo Badajoz (Spain) (Portugal) ComarcaComarca de BadajozMonte deMonte dosMonte BadajozMonte Arneirosdos dos Arneirosdos Arneiros Holm Arneiros oak, Holm eucalyptus, Corkoak, Cork oak, eucalyptus, scrub Cork Cork oak,scr uboak, oak,scr ubscr scr ub scrub ub Montemor-o-Novo Badajoz Montemor-o-Novo Montemor-o-Novo Montemor-o-Novo(Spain) Badajoz (Portugal)for (Portugal) (Spain)triangular (Portugal) (Portugal) analysis with an electronic nose. Some individuals were assigned to perform a GC–MS Monte dos ArneirosSierra Suroeste Chestnut, Cork oak, holm scrub oak, cork oak, scrub Montemor-o-Novo Badajoz (Spain) (Portugal) Monte dosMonte Arneiros dosSierra ArneirosSierra SuroesteSierraSierra Suroeste Suroeste Suroeste Cork Chestnut, Chestnut,oak, Chestnut, Chestnut, holm Corkscrub holm oak, oak, holm holm oak,cork scruboak, oak, corkoak, cork cork oak,scrub oak, oak,scrub Montemor-o-Novo scrub scrub Montemor-o-Novo Badajoz Badajoz (Portugal) Badajoz (Spain) Badajoz (Spain)analysis (Spain) (Spain) (Portugal) of all mushroom species participating in the study. SierraSierraSierra SuroesteSuroeste Suroeste Chestnut, Chestnut, Chestnut, holmholm holm oak,oak, oak, corkcork cork oak,oak, oak, scrubscrub scrub Badajoz Badajoz Badajoz (Spain)(Spain) (Spain) SierraFor the Suroeste tests, the hats [31] Chestnut, from individuals holmoak, of optimal cork oak, mushrooms scrub were chosen. BadajozIn a period (Spain) of For FortheFor tests,theFor the tests,the the tests, tests, hatsthe the hats the[31] hats hats [31]from [31] [31]from individuals from from individuals individuals individuals of optimal of optimalof of optimal optimalmushrooms mushrooms mushrooms mushrooms were were chosen. were were chosen. chosen. Inchosen.2.2. a Inperiod GC–MS aIn Inperiod a a periodof periodMeasurements of of of time not greater than four hours from the harvest, they were put under a process of freezing to −24 For the tests,time thetime hatsnot not greater [31] greater from than than individualsfour four hours hours from of from optimal the the harvest, harvest, mushrooms they they were were were put put under chosen. under a process a In process a period of freezingof freezing of to −to24 − 24 ForFor the the °C.tests, tests,time Prior thenot the to time hatsgreater hatsfreezing, not[31] [31] thangreater from theyfrom four thanindividualswere individualshours four quartered, from hours ofthe of fromoptimal andharvest,optimal theeach harvest,mushroomsthey mushroomssample were they was put werewere weighed underwere put chosen. chosen.a underprocessand storedIn a In ofprocessa aperiodfreezing periodin coded of In of freezing to oforder − 24 to to verify −24 the existence of differential volatile compounds in each of the 6 species of mushrooms studied in this work, some analyses had previously been carried out using 0.5 g of each timetimetime2.2. notnot not GC–MS greatergreater greaterodorless°C. Measurementsthan than °C.Priorthan bags.°C. Prior °C.fourfour fourtoPrior The Prior freezing, hourstohours hours majortofreezing, to freezing, fromfreezing,from partsfromthey they thethewere were thethey they harvest,harvest, were harvest,allocated quartered,were were quartered, theyquartered,they forquartered,they trand werewereiangular were andeach and putput andputeach sensorysample eachunder under eachunder sample sample analysis samplewas aa a processprocess wasprocessweighed was andwasweighed of weighedof the weighedof freezingandfreezing remainingfreezing andstored and andstored toto tostoredparts in −stored 24 −coded 24in incoded in coded coded °C.°C.°C. PriorPrior Prior toto to for freezing,freezing, odorlessfreezing, triangularodorlessodorlessodorless bags.theythey theyanalysis bags. The werebags.were werebags. Themajor with Thequartered, quartered, Themajorquartered, anparts major major electronic parts were parts parts wereandand allocatedand werenose. were allocatedeacheach each allocatedSome allocated forsamplesample sample tr individualsforiangular fortr for iangular waswastr wastriangulariangular sensory wereweighedweighed weighed sensory assignedsensory sensoryanalysis analysisandand and analysis to analysis and stored stored performstored andthe and remaining and theinin ain the GC–MSremaining codedthecoded coded remaining remaining parts parts parts parts for triangularfor fortriangularfor triangular triangular analysis analysis analysis analysiswith with an with electronicwith an anelectronic an electronic electronic nose. nose. Some nose. nose. Some individuals Some Some individuals individuals individuals were were assigned were were assigned assigned assigned to perform to performto to perform perform a GC–MS a GC–MS a aGC–MS GC–MS odorlessodorlessIn bags. bags. orderanalysis The The tomajor ofmajor verifyall mushroomparts parts the were were existencespecies allocated allocated participating for of for tr di triangulariangularffiangular inerential the study. sensorysensory sensory volatile analysisanalysis analysis compounds andand and thethe the remainingremaining remaining in each partsparts parts of the 6 species of analysisanalysisanalysis analysisof all of mushroom allof of allmushroom all mushroom mushroom species species species participatingspecies participating participating participating in the in study. thein in the study.the study. study. forforformushrooms triangulartriangular triangular analysisanalysis analysis studied withwith with in anan thisan electronicelectronic electronic work, nose. somenose. nose. SomeSome analysesSome individualsindividuals individuals had previously werewere were assignedassigned assigned been toto to performperform carried perform aa out a GC–MSGC–MS GC–MS using 0.5 g of each of 2.2. GC–MS Measurements analysisanalysisthem. of Theof all all mushroom2.2. technique mushroom GC–MS2.2.2.2. GC–MS2.2. GC–MS GC–MSMeasurements species solid-phasespecies Measurements Measurements Measurements participating participating microextraction in in the the study. study. followed by gas chromatography–mass spectrometry In order to verify the existence of differential volatile compounds in each of the 6 species of (SPME-GC–MS) wasIn InorderIn applied.order order to verifyto to verify verify A the GC–MS the existencethe existence existence instrument of differentialof of differential differential manufactured volatile volatile volatile compound compound compound bys BRUKERins seachin in each each of theof (Billerica,of the 6the species 6 6species species MA,of of of United 2.2.2.2. GC–MS GC–MS Measurementsmushrooms MeasurementsIn order studied to verify in this the work, existence some analysesof differential had previously volatile compoundbeen carrieds inout each using of 0.5the g 6of species each of States) wasmushrooms usedmushroomsmushroomsmushrooms for studied this studied purpose.studied instudied this in thiswork,in in this Anthiswork, somework, work, 85- some µanalyses somem some analyses Carboxen analyses analyses had hadpreviously had hadpreviously/Polydimethylsiloxane previously previously been been carried been been carried carried outcarried usingout out usingout (CAR0.5 using using g0.5 of /0.5 g PDMS)each0.5 of g geachof of each each fiber was InIn orderorder toto verifyverify thethe existenceexistence of differential volatile compoundss inin eacheach ofof thethe 66 speciesspecies ofof exposedIn order to theto verify headspace, the existence while of the differential sample wasvolatile shaken compound at 40s◦ inC foreach 30 ofmin. the 6 Subsequently,species of desorption mushrooms studied in this work, some analyses had previously been carried out using 0.5 g of each mushroomstook place studied in the in injection this work, port some foranalyses 10 min. had previously The identification been carried of out the using compounds 0.5 g of each was done based

on the results provided by the National Institute of Standards and Technology(NIST) spectroscopy database [32]. The temperature program used consisted of three phases. First, the oven temperature was set at 45 ◦C for 1 min. It was then increased to 200 ◦C at a rate of 5 ◦C/min. Finally, it increased by 20 ◦C /min until it reached 250 ◦C, a temperature which was maintained for 5 min. In addition, between the different injections, a cleaning procedure of the SPME fiber was carried out. To do this, the SPME fiber was kept at 300 ◦C for 30 min.

2.3. Triangular Sensory Test Discriminatory sensory tests are those in which it is not intended to know the subjective sensation produced by the individual tested, but to establish the differences that may exist between two or more samples [33]. In this study, a particular triangular sensory test was carried out. In this test, in addition to setting a level of significance to validate the statistical differentiation between the odors of the samples, it was also intended to obtain more specific probabilistic data for each number of correct answers. In this way, the results could be compared to those obtained in the tests with the electronic nose. Foods 2019, 8, 414 5 of 20

Four sessions were held over a month, with a total of 16 combinations or triangles (4 per session), in which all the participating species were compared to each other (Table3). In each triangle, three samples were presented simultaneously. They were chosen from among the six species under study, of which two belonged to the same mushroom and the remaining one to a diﬀerent one. In all the tests, 25 (not trained) consumer judges participated forming a very heterogeneous group in mycological knowledge (experts, amateurs, students, curious people ... ), gender and age. All of them attended the Jornadas Micológicas 2018, held during the month of November at Centro de Profesores y Recursos in Badajoz (Spain).

Table 3. Sensorial triangle codes for each session.

Triangles Session ABCD Day 1 D1-A_MPP D1-B_CaCaM D1_C_PhCaPh D1-D_RCaCa Day 2 D2-A_RMR D2-B_PRP D2-C_PCC D2-D_MMC Day 3 D3-A_PhPPh D3-B_RRC D3-C_CCPh D3-D_CPP Day 4 D4-A_CaPCa D4-B_MMPh D4-C_CaCC D4-D_PhPhR R: A. rubescens; P: A. pantherina; C: A. citrina; M: A. muscaria; Ph: A. phalloides; Ca: A. caesarea.

After smelling the mushrooms, the judges were asked to mark, on the test sheet provided, the number corresponding to the one they considered diﬀerent from each of the triangles of the session (Figure2). In order to avoid possible stimulus errors, equal odorless containers were used, which made it impossible to recognize the mushrooms by any other sense than smell. The answers were Foodscoded 2019 with, 8, 414 four-digit random numbers to avoid errors of expectation [16]. 6 of 19

Figure 2. JudgeJudge from from the the sensory sensory panel panel participat participatinging in the triangular sensory test.

2.4. Electronic Electronic Nose A home-developedhome-developed low-cost low-cost electronic electronic nose nose was usedwas forused measurements. for measurements. The Wireless The ElectronicWireless ElectronicNose (WINOSE Nose 6)(WINOSE (Figure3 )6) was (Figure designed 3) was and desi developedgned and jointly developed by the jointly Spanish by Council the Spanish for Scientiﬁc Council forResearch Scientific (CSIC, Research Madrid, (CSIC, Spain) Madrid, and the Sp Universityain) and the of University Extremadura. of Extremadura. The electronic nose was divided into the following functional parts: an aroma extraction system, a detection system or sensor, a control and instrumentation system, and a data and signal processing system. A block diagram of the WINOSE 6 is shown in Figure 4.

Figure 3. Curve of the response of the sensors to the D1-A_MPP triangle test.

Foods 2019, 8, 414 6 of 19

Figure 2. Judge from the sensory panel participating in the triangular sensory test.

2.4. Electronic Nose A home-developed low-cost electronic nose was used for measurements. The Wireless Electronic Nose (WINOSE 6) (Figure 3) was designed and developed jointly by the Spanish Council for Scientific Research (CSIC, Madrid, Spain) and the University of Extremadura. The electronic nose was divided into the following functional parts: an aroma extraction system, a detection system or sensor, a control and instrumentation system, and a data and signal processing Foods 2019, 8, 414 6 of 20 system. A block diagram of the WINOSE 6 is shown in Figure 4.

Figure 3. CurveCurve of of the the response response of the sens sensorsors to the D1-A_MPP triangle test.

The electronic nose was divided into the following functional parts: an aroma extraction system, a detection system or sensor, a control and instrumentation system, and a data and signal processing Foodssystem. 2019, A 8, block414 diagram of the WINOSE 6 is shown in Figure4. 7 of 19

Figure 4. Block diagram of the electronic nose. Figure 4. Block diagram of the electronic nose. It had an array of 8 interchangeable sensors that were controlled by a dsPIC33FJ128GP306 It had an array of 8 interchangeable sensors that were controlled by a dsPIC33FJ128GP306 microcontroller from Microchip. In addition, the electronic nose contained integrated controls and microcontroller from Microchip. In addition, the electronic nose contained integrated controls and instrumentation electronics, rechargeable batteries, a touch screen, and an IEEE 802.11 transceiver for instrumentation electronics, rechargeable batteries, a touch screen, and an IEEE 802.11 transceiver for wireless communications. The e-nose was equipped with an Liquid Cristal Display (LCD) touch screen wireless communications. The e-nose was equipped with an Liquid Cristal Display (LCD) touch where the responses of the sensors and the main measurement parameters were shown. screen where the responses of the sensors and the main measurement parameters were shown. The software that controlled the entire process was developed with the NI LabVIEW program The software that controlled the entire process was developed with the NI LabVIEW program from National Instruments. It made it possible to preset all of the parameters of the analysis (e.g., the from National Instruments. It made it possible to preset all of the parameters of the analysis (e.g., the sensor heater temperatures, pump power, or absorption and desorption times). sensor heater temperatures, pump power, or absorption and desorption times). The metal oxide (MOX) gas sensors implemented in the system, which have already been successfully used in the recognition of fungi, are shown in Table 4, which indicates the manufacturers and the preset temperature for the heaters. The MiCS4514 was composed of two sensors that were included in the same package, while the MiCS6814 was composed of three. The rest included only one sensor per chip. The sensors were located inside a cell connected to a three-way solenoid valve (SMC S70_ES), which permitted selection between two inputs: one for the sample to be measured and the other one for a reference gas (in our case free air), which was passed through an active carbon filter to obtain the reference baseline. The electronic nose also had relative humidity and temperature sensors in the same package (Sensirion SHT21, Sensirion, Stäfa, Switzerland) and a micropump (RietschleThomas model 2002, Gardner Denver, Milwaukee, WI, USA).

Table 4. Sensors used by WINOSE 6.

V heaters (V) Sensor Manufacturer Model Minimum Maximum V Configured S1 Cambridge CCS801 1.2 1.8 1.8 S2 Cambridge CCS803 1.2 1.8 1.8 S3 SGX MICS4514 OX 1.3 1.9 1.7 S4 SGX MICS4514 RED 2.2 2.5 2.4 S5 Figaro TGS8100 1.8 1.8 1.8 S6 SGX MICS6814 OX 1.3 1.9 1.7 S7 SGX MICS6814 RED 2.2 2.6 2.4 S8 SGX MICS6814 NH3 2 2.4 2.2 Note: Sensors S3 and S4 and S6, S7, and S8 were in the same encapsulation.

Foods 2019, 8, 414 7 of 20

The metal oxide (MOX) gas sensors implemented in the system, which have already been successfully used in the recognition of fungi, are shown in Table4, which indicates the manufacturers and the preset temperature for the heaters. The MiCS4514 was composed of two sensors that were included in the same package, while the MiCS6814 was composed of three. The rest included only one sensor per chip. The sensors were located inside a cell connected to a three-way solenoid valve (SMC S70_ES), which permitted selection between two inputs: one for the sample to be measured and the other one for a reference gas (in our case free air), which was passed through an active carbon ﬁlter to obtain the reference baseline. The electronic nose also had relative humidity and temperature sensors in the same package (Sensirion SHT21, Sensirion, Stäfa, Switzerland) and a micropump (RietschleThomas model 2002, Gardner Denver, Milwaukee, WI, USA).

Table 4. Sensors used by WINOSE 6.

V heaters (V) Sensor Manufacturer Model Minimum Maximum V Conﬁgured S1 Cambridge CCS801 1.2 1.8 1.8 S2 Cambridge CCS803 1.2 1.8 1.8 S3 SGX MICS4514 OX 1.3 1.9 1.7 S4 SGX MICS4514 RED 2.2 2.5 2.4 S5 Figaro TGS8100 1.8 1.8 1.8 S6 SGX MICS6814 OX 1.3 1.9 1.7 S7 SGX MICS6814 RED 2.2 2.6 2.4 S8 SGX MICS6814 NH3 2 2.4 2.2 Note: Sensors S3 and S4 and S6, S7, and S8 were in the same encapsulation. Foods 2019, 8, 414 8 of 19

2.4.1. Sampling Method Static headspace headspace with with a atransfer transfer of ofeffluent eﬄuent technique technique was was used used for volatile for volatile extraction extraction [34]. Each [34]. Eachpiece pieceof sample of sample to be toanalyzed, be analyzed, after being after being thawed, thawed, was placed was placed in a 20-mL in a 20-mL glass glassvial and vial sealed and sealed with witha septum a septum and andscrew screw cap capto allow to allow the theinjectors injectors to toenter. enter. The The vials vials were were immersed immersed in in a a thermostatic bath at 30 ◦°CC (Figure5 5).). InIn orderorder toto resetreset thethe samesame positionposition (the(the inclinationinclination andand penetrationpenetration ofof thethe injection needles) in allall analyses,analyses, aa capcap withwith stopstop guidesguides waswas designeddesigned andand used.used.

Figure 5. TestTest of mushrooms with the electronic nose WINOSE 6.

The analysis began with progressive heating of the sensor surfaces, and once they reached their stabilized reference temperatures (Table(Table4 4),), the the programmed programmed tasks tasks began. began. EachEach datadata acquisitionacquisition cyclecycle consisted of two parts. First was was a a 7-min 7-min desorption desorption period, period, during which filtered ﬁltered free air was injected into the cell where the sensors were located. Every second, the system took a reading of the resistive value supplied by each sensor, thus obtaining the reference baselines. After that period of time, the valve that allowed the absorption of volatiles of the sample to be analyzed was activated for 1 min, taking one reading per second. In order to make the experiment as similar as possible to the triangular sensory test, 3 absorptions were made for each mushroom sample, and a filtered free air absorption was always made before, between, and after each sample (Figure 3).

2.4.2. Preprocessing To select the descriptive parameters of the resistive values obtained by the sensors, the fractional baseline manipulation procedure was used. It consisted of dividing the difference between the resistive value obtained at the reference baseline and the response of each sensor to the volatiles emanating from the corresponding mushroom sample by the reference baseline (Equation (1)). In this way, the multiplicative drift was eliminated, generating a dimensionless and normalized response [35]:

𝑅 −𝑅 𝑅𝑖 = (1) 𝑅

Foods 2019, 8, 414 8 of 20 into the cell where the sensors were located. Every second, the system took a reading of the resistive value supplied by each sensor, thus obtaining the reference baselines. After that period of time, the valve that allowed the absorption of volatiles of the sample to be analyzed was activated for 1 min, taking one reading per second. In order to make the experiment as similar as possible to the triangular sensory test, 3 absorptions were made for each mushroom sample, and a ﬁltered free air absorption was always made before, between, and after each sample (Figure3).

2.4.2. Preprocessing To select the descriptive parameters of the resistive values obtained by the sensors, the fractional baseline manipulation procedure was used. It consisted of dividing the diﬀerence between the resistive value obtained at the reference baseline and the response of each sensor to the volatiles emanating from the corresponding mushroom sample by the reference baseline (Equation (1)). In this way, the multiplicative drift was eliminated, generating a dimensionless and normalized response [35]:

Rre f Raroma Ri = − (1) Rre f

2.5. Statistical Analysis The objective of this work was to compare the discriminative capacity (between the analyzed samples) that the panelists and the electronic nose had. To this end, analogous guidelines were maintained in the two procedures, both in terms of data retrieval, as commented on previously, and in terms of statistical analysis. According to these criteria and given that the participating judges did not have previous training, it was considered that the device results should be obtained through the use of an unsupervised learning method. In addition, this was all the more convenient as there were very few readings for each triangle, which greatly limited the training process. In order to validate the statistical accuracy in the discrimination of the analyzed mushrooms, a level of signiﬁcance was established in both cases, α = 5%, which is the level usually used in this type of test [36]. As this was a comparative study, the level of statistical conﬁdence was also assessed (1–α), which was derived from the number of correct answers of each triangle analyzed by means of the calculation of the probabilistic value p-value.

2.5.1. Sensory Analysis The null hypothesis in a triangular test establishes that the probability of randomly choosing a different sample is p = 1/3 [37]. A null hypothesis (H0) was established: if we odorously compared two different mushrooms, A and B, of the genus Amanita of the species A. caesarea, A. phalloides, A. rubescens, A. pantherina, A. muscaria, or A. citrina, which were collected from different points of the southwest of the Iberian Peninsula during the month of November 2018, we would not find odorous differences between them (Equation (2)). If H0 were rejected, we would not reject the alternative hypothesis H1 (Equation (3)), which would give credibility to the existence of odorous differences between the individuals analyzed: H0 : µA = µB (2)

H1 : µA , µB (3) In a first approximation, the tables published by Reference [38], which were based on a binomial distribution, were used. There, the minimum number of correct answers was indicated, depending on the number of participating judges, for establishing significant differences between the two samples with the prefixed level of significance α. Second, in order to compare this to the results obtained by the electronic nose, the p-value of each triangle was calculated as a function of the number of correct answers. Equation (4) was used, together with Equation (5), where the probabilities of a higher number of correct answers than obtained were assessed. Table5 indicates the p-value obtained for a certain Foods 2019, 8, 414 9 of 20

number of correct answers by 25 judges, indicating the rejection of H0 with α = 0.5 and also the limits for α = 5%, α = 1%, and α = 0.1%: ! n y n y P(y) = p (1 p) − (4) y −

p-value = P(y) + P(y + 1) + + P(n) (5) ··· where n = 25 and p = 1/3 at y = no. of correct answers.

Table 5. Calculation of the p-value for a hypothesis contrast using the binomial distribution and comparison to the chosen level of signiﬁcance (α = 0.5).

n y p-Value Reject H0 α (%) 25 10 0.1264 NO >5% 25 11 0.0862 NO >5% 25 12 0.0503 NO >5% 25 13 0.0251 YES <5% 25 14 0.0108 YES <5% 25 15 0.0040 YES <1% 25 16 0.0012 YES <1% 25 17 0.0003 YES <0.1% 25 18 0.0001 YES <0.1% 25 19 0.0000 YES <0.1%

n = number of panelists; y = right answers; H0 = null hypothesis; α = level of signiﬁcance.

2.5.2. Electronic Nose Data Once the data supplied by the sensors were preprocessed, the fingerprints of the volatile organic compounds (VOCs) [39] present in each mushroom were obtained. As defined by Reference [40], the term volatile organic compounds (VOCs) refers to those organic compounds that are present in the atmosphere as gases but that would be liquids or solids under normal conditions of temperature and pressure. In this same publication [40], the enhancement of the sensing performances of MOX gas sensors for detecting toxic gases and VOCs is also highlighted. As has been mentioned, mushrooms generate a great diversity of these VOCs [6]. With these data, after averaging, the aroma obtained from each triangle was represented by means of radial graphs (see Figure6). The automatic classification of these fingerprints was performed using the multivariate statistical agglomerative method hierarchical cluster analysis (HCA) [41]. The HCA refers to an unsupervised learning technique that aims to find patterns or clusters within a set of observations. The partitions are established in such a way that the observations that are within the same group are similar between them and different from those of other groups. For the hierarchical cluster analysis, the hcluster function (https://www.rdocumentation.org/packages/amap/versions/0.8-16/topics/hcluster), which belongs to the statistical analysis program R (https://www.r-project.org/) was used. Before the analysis, a series of tests was carried out to determine the most suitable methods for the analysis of the data obtained, as this is not a trivial matter and should be checked when an HCA is applied [41]. From these checks, the Euclidean method was chosen for distances and the single method [42] for clusters (Figure7). Foods 2019, 8, 414 10 of 19

Once the data supplied by the sensors were preprocessed, the fingerprints of the volatile organic compounds (VOCs) [39] present in each mushroom were obtained. As defined by Reference [40], the term volatile organic compounds (VOCs) refers to those organic compounds that are present in the atmosphere as gases but that would be liquids or solids under normal conditions of temperature and pressure. In this same publication [40], the enhancement of the sensing performances of MOX gas sensors for detecting toxic gases and VOCs is also highlighted. As has been mentioned, mushrooms generate a great diversity of these VOCs [6]. With these data, after averaging, the aroma obtained from each triangle was represented by means of radial graphs (see Figure 6). The automatic classification of these fingerprints was performed using the multivariate statistical agglomerative method hierarchical cluster analysis (HCA) [41]. The HCA refers to an unsupervised learning technique that aims to find patterns or clusters within a set of observations. The partitions are established in such a way that the observations that are within the same group are similar between them and different from those of other groups. For the hierarchical cluster analysis, the hcluster function (https://www.rdocumentation.org/packages/amap/versions/0.8-16/topics/hcluster), which belongs to the statistical analysis program R (https://www.r-project.org/) was used. Before the analysis, a series of tests was carried out to determine the most suitable methods for the analysis of the data obtained, as this is not a trivial matter and should be checked when an HCA is applied [41]. From these checks, the Euclidean method was chosen for distances and the single method [42] for Foods 2019, 8, 414 10 of 20 clusters (Figure 7).

Foods 2019, 8, 414 11 of 19 Figure 6. GraphicalGraphical representation representation of the aromas obtained from the triangle D2-B_PRP.

Figure 7. 7. DifferencesDifferences in dendrograms in dendrograms for the for same the data same with data average with (A average) and single (A) (B and) grouping single (methods.B) grouping methods. In order to obtain the p-value of each grouping, the R pvcluster library was used [43] with a In order to obtain the p-value of each grouping, the R pvcluster library was used [43] with a significance level of α = 0.5. significance level of α = 0.5. Once the dendrogram was created, it had to be validated, assessing to what extent its structure Once the dendrogram was created, it had to be validated, assessing to what extent its structure reflected the original distances between observations. One of the most commonly used criteria for reflected the original distances between observations. One of the most commonly used criteria for that is the cophenetic correlation coefficient of the dendrogram [44]. If a value close to the unit is that is the cophenetic correlation coefficient of the dendrogram [44]. If a value close to the unit is obtained, this means that there is a good hierarchical structure between the elements analyzed. Some obtained, this means that there is a good hierarchical structure between the elements analyzed. Some publications have indicated that a good partition must have a cophenetic correlation of at least 0.85 [45]. publications have indicated that a good partition must have a cophenetic correlation of at least 0.85 [45].

3. Results and Discussion

3.1. GC–MS The values shown in Table 6 for each compound are absolute values of measured areas, so they do not correspond to a certain concentration. Some compounds are not detected (ND)

Table 6. Volatile compounds identified by gas chromatography–mass spectrometry (GC–MS) in the mushroom species analyzed.

Samples Compound A. A. A. A. rubescens A. citrina A. caesarea pantherina muscaria phalloides 3-Methyl-butanal 27 ND ND 13 402 ND 2-Methyl-butanal 87.4 ND 0 ND 1300 ND Acetic acid ND ND ND ND ND ND Acetoin ND ND ND ND ND ND 3-Methyl-1-butanol ND ND 10.9 15.8 507 11.7 1-Octanol ND ND ND ND ND ND Hexanal 20.5 940 6.42 28 26.3 13.9 Styrene 18.6 17.9 38.8 70.8 55.1 7.56 Heptanal ND 4.48 ND ND ND ND Benzaldehyde ND ND ND ND 24.6 ND 1-Hept-3-one ND 6.68 ND 84.5 ND ND 1-Octen-3-ol 18.4 308 10.1 78.7 3.44 363 3-Octanone 159 209 280 917 153 392 3-Octanol 12.3 28.4 ND ND ND 19.7

Foods 2019, 8, 414 11 of 20

3. Results and Discussion

3.1. GC–MS The values shown in Table6 for each compound are absolute values of measured areas, so they do not correspond to a certain concentration. Some compounds are not detected (ND).

Table 6. Volatile compounds identiﬁed by gas chromatography–mass spectrometry (GC–MS) in the mushroom species analyzed.

Samples Compound A. A. A. rubescens A. citrina A. muscaria A. caesarea pantherina phalloides 3-Methyl-butanal 27 ND ND 13 402 ND 2-Methyl-butanal 87.4 ND 0 ND 1300 ND Acetic acid ND ND ND ND ND ND Acetoin ND ND ND ND ND ND 3-Methyl-1-butanol ND ND 10.9 15.8 507 11.7 1-Octanol ND ND ND ND ND ND Hexanal 20.5 940 6.42 28 26.3 13.9 Styrene 18.6 17.9 38.8 70.8 55.1 7.56 Heptanal ND 4.48 ND ND ND ND Benzaldehyde ND ND ND ND 24.6 ND 1-Hept-3-one ND 6.68 ND 84.5 ND ND 1-Octen-3-ol 18.4 308 10.1 78.7 3.44 363 3-Octanone 159 209 280 917 153 392 3-Octanol 12.3 28.4 ND ND ND 19.7 Octanal ND 11.4 ND 9.44 ND ND 2-Ethyl-1-hexanol 40 105 ND 29.4 ND 10.91 2-Octenal ND 2.19 ND ND ND ND Borneol ND ND ND ND ND ND

Nowadays, it is generally accepted that compounds containing eight carbons are the primary volatile contributing to the taste of fungi. Compounds of this class include 3-octanol, 3-octanone, 1-octen-3-ol, etc. [46]. It can be observed that they appeared in diﬀerent proportions in the chromatographic analysis carried out, which indicated a possible odorous variation in each of the species analyzed.

3.2. Sensory Analysis Table7 shows the results of the triangular sensory analysis for a significance level of α = 0.5 expressed as a function of the confidence level (1–α). The p-value corresponding to each triangle is enclosed in parentheses, depending on the number of correct answers. From these results, it can be concluded that in over 87% of cases, H0 could be rejected and H1 accepted. When the null hypothesis H0 is rejected, there is statistical evidence that the alternative hypothesis H1 is acceptable. On the other hand, if it is not rejected, there is no statistical evidence to accept it. Because the p-value of the triangles in which H0 was accepted was very close to α, a type II error was likely to have occurred. In other words, the null hypothesis was accepted even though it was false. In both cases, it was very possible that the sensory test did not have sufficient contrast intensity to determine the stipulated percentage of odor difference. This could be solved by increasing the number of judges [47]. Foods 2019, 8, 414 12 of 20

Table 7. Results of the triangular sensory analysis for a signiﬁcance level of α = 0.5 and p-value for each triangle.

Conﬁdence Level 95% >95% >99% >99.9% ≤ Null Hypothesis Accepted Rejected Rejected Rejected Number of Trials 2 (12.5%) 6 (37.5%) 3 (18.75%) 5 (31.25%) D3-A_PhPPh D1-B_CaCaM D1-A_MPP D1-C_PhCaPh (91.38%) (98.92%) (99.88%) (99.97%) D4-A_CaPCa D1-_RcaCa D3-B_RRC D2-B_PRP (94.97%) (97.49%) (99.60%) (100%) D2-A_RMR D4-A_CaPCa D3-D_CPP (98.92%) (99.88%) (100%) Triangles D2-C_PCC D4-D_PhPhR D4-B_MMPh (97.49%) (99.60%) (100%) D2-D_MMC D4-C_CaCC (97.49%) (100%) D3-C_CCPh (98.92%)

To sum up, it could be statistically affirmed that we humans are capable of differentiating between mushrooms through smell, and therefore, we can begin to answer affirmatively the first of the questions presented by this research. A further interesting aspect to observe is how sensory results improve over time. Perhaps it was due to the increased motivation shown by the participating judges during the development of the trials and because of the eventual learning from the tasting practice.

3.3. Electronic Nose Results

3.3.1. Graphical Results The response curve of each sensor, which was obtained directly from the e-nose, anticipated the possible volatile differences between the two types of mushrooms analyzed in each triangle. These differences were more accurately visualized once preprocessed and averaged, as shown in Figure8, where the radial graph is shown next to the dendrogram of the triangles D1-A_MPP, D1-C_PhCaPh, D2-A_RMR, and D3-C_CCPh. The sensors used, despite having a long recovery time (desorption period), showed a fairly stable performance. It should be noted that the e-nose tests were performed under very similar humidity and temperature conditions. Cyclically, calibrations were carried out with 10% ethanol, in which good temporal repeatability in the responses was observed. It was also observed that some sensors had a wider response margin than others. Some of the previous considerations are also reflected in Figure8, where, for instance, we can observe that the sample of Amanita muscaria from the D-1_MPP test was very similar to the one obtained from the D2-A_RMR test. Foods 2019, 8, 414 13 of 19

The response curve of each sensor, which was obtained directly from the e-nose, anticipated the possible volatile differences between the two types of mushrooms analyzed in each triangle. These differences were more accurately visualized once preprocessed and averaged, as shown in Figure 8, where the radial graph is shown next to the dendrogram of the triangles D1-A_MPP, D1-C_PhCaPh, D2-A_RMR, and D3-C_CCPh. The sensors used, despite having a long recovery time (desorption period), showed a fairly stable performance. It should be noted that the e-nose tests were performed under very similar humidity and temperature conditions. Cyclically, calibrations were carried out with 10% ethanol, in which good temporal repeatability in the responses was observed. It was also observed that some sensors had a wider response margin than others. Some of the previous considerations are also reflected in Figure 8, where, for instance, we can observe that the sample of FoodsAmanita2019, 8, muscaria 414 from the D-1_MPP test was very similar to the one obtained from the D2-A_RMR13 of 20 test.

(A) (B)

Figure 8. Cont.

Foods 2019, 8, 414 14 of 20 Foods 2019, 8, 414 14 of 19

(E) (F)

(G) (H)

FigureFigure 8. 8.Odor Odor ﬁngerprintsfingerprints and and dendrogram dendrogramss of ofthe the triangles: triangles: D1-A_MPP D1-A_MPP (A) (B (A), )(D1-C_PhCaPhB), D1-C_PhCaPh (C, (CD,D),), D2-A_RMR D2-A_RMR (E ()E ()(F),F ),and and D3-C_CCPh D3-C_CCPh (G) ( G(H)().H ).

3.3.2.3.3.2. Results Results Obtained Obtained through through thethe MultivariateMultivariate Statistical Statistical Method Method HCA HCA TableTable8 shows 8 shows the the results results of all of analyses all analyses performed. performed. It was It determinedwas determined that onlythat twoonly oftwo the of sensory the trialssensory (D3-A_PhPPh trials (D3-A_PhPPh and D4-A_CaPCa) and D4-A_CaPCa) and one and of theone electronic of the electronic nose trials nose (D4-A_CaPCa)trials (D4-A_CaPCa) did not rejectdid thenot nullreject hypothesis the null hypothesis at the established at the established signiﬁcance significance level. level.

Table 8. Comparative results of sensory analysis and electronic nose analysis. Negative values in the “% Difference” column indicate a better response from the electronic nose. Cophenetic correlation coefficient values.

Humans versus Machines: Triangular Analysis Cophenetic Triangle Sensory Analysis Electronic Nose % Difference Coeff. D1-A_MPP 0.9988 0.9949 0.9979 0.39% D1-B_CaCaM 0.9892 1.0000 0.8200 −1.08% D1-C_PhCaPh 0.9997 1.0000 0.9789 −0.03% D1-D_RcaCa 0.9749 1.0000 0.9508 −2.51% D2-A_RMM 0.9892 1.0000 0.9244 −1.08%

FoodsFoods2019 2019, 8, 414, 8, 414 15 15of of19 20

D2-B_PRP 1.0000 1.0000 0.7322 0.00% Table 8.D2-C_PCCComparative results0.9749 of sensory analysis and 1.0000 electronic nose 0.8064analysis. Negative− values2.51% in the “% DiD2-D_MMCfference” column indicate0.9749 a better response 0.9948 from the electronic 0.7585 nose. Cophenetic−1.99% correlation coefficientD3-A_PhPPh values. 0.9138 0.9970 0.9640 −8.32% D3-B_RRC 0.9960 1.0000 0.8782 −0.40% D3-C_CCPh Humans0.9892 versus Machines: 1.0000 Triangular Analysis 0.9271 −1.08% D3-D_CPP 1.0000 1.0000 0.8064 0.00% Triangle Sensory Analysis Electronic Nose Cophenetic Coeff. % Difference D4-A_CaPCa 0.9497 0.8853 0.9629 6.44% D1-A_MPPD4-B_MMPh 0.99881.0000 0.9949 0.9981 0.99790.9622 0.19% 0.39% D1-B_CaCaMD4-C_CaCC 0.98921.0000 1.0000 0.9738 0.82000.9992 2.62%1.08% − D1-C_PhCaPh 0.9997 1.0000 0.9789 0.03% D4-D_PhPhR 0.9960 1.0000 0.8913 −0.40%− D1-D_RcaCa 0.9749 1.0000 0.9508 2.51% − D2-A_RMM 0.9892 1.0000 0.9244 1.08% These results evidence that, in most trials, the electronic nose had a high− rate of odor D2-B_PRP 1.0000 1.0000 0.7322 0.00% discrimination between samples. Over 80% of the triangles had a confidence level above 99%, D2-C_PCC 0.9749 1.0000 0.8064 2.51% − indicatingD2-D_MMC that, potentially, 0.9749 a lower significance 0.9948 level could have been 0.7585 applied if wider sampling1.99% had − beenD3-A_PhPPh performed. The radial 0.9138 graph (A) and the grouping 0.9970 dendrogram 0.9640 (B) of the triangle D3-A_PhPPh,8.32% − whichD3-B_RRC was one of the triangles 0.9960 exceeding 99%, 1.0000 are shown in Figure 0.8782 9. Notice that the0.40% graphical − differenceD3-C_CCPh is perfectly in 0.9892accordance with the 1.0000 hierarchical grouping. 0.9271 The data obtained1.08% by the − hierarchicalD3-D_CPP clustering analysis 1.0000 were evaluated 1.0000 (Table 8) using the 0.8064cophenetic correlation 0.00% coefficient method.D4-A_CaPCa 0.9497 0.8853 0.9629 6.44% D4-B_MMPh 1.0000 0.9981 0.9622 0.19% The D4-A_CaPCa triangle was the only one that did not reject the null hypothesis, with an D4-C_CaCC 1.0000 0.9738 0.9992 2.62% associatedD4-D_PhPhR p-value of 0.1147. 0.9960 In the radial graph 1.0000 of Figure 10A, the discrimination 0.8913 between0.40% both types of mushrooms was not clear. This was confirmed by the detection of more than two− groups by the hcluster function, which evidenced the impossibility of differentiating one mushroom from another byThese grouping, results as evidence observed that, in Figure in most 10B. trials, In Table the electronic 6, where nosethe aroma had a compounds high rate of odorof mushrooms discrimination are betweenindicated, samples. it can be Over seen 80% that ofA. thecaesarea triangles (Ca) and had A. a pantherina confidence (P)level had the above closest 99%, profile. indicating This may that, potentially,suggest that a lower there significance are similarities level between could have the vo beenlatile applied organic if widercompounds sampling that hademerge been from performed. them. TheHowever, radial graph in contrast, (A) and in the the grouping corresponding dendrogram sensory (B) trial, of thea high triangle level of D3-A_PhPPh, discrimination which was obtained. was one of the trianglesDespite exceeding this particular 99%, are situation, shown init Figurecould be9. Noticestatistically that theaffirmed graphical that ditheff electronicerence is perfectlynose was in accordanceable to differentiate with the hierarchical between the grouping. vast majority The of data the obtainedanalyzed bymushrooms the hierarchical with enough clustering clarity. analysis This wereanswers evaluated affirmatively (Table8) using the second the cophenetic question correlation presented: coe machinesfficient method.can help us to find odorous differences in mushrooms.

(A) (B)

FigureFigure 9. Radial9. Radial graph graph (A (A) and) and grouping grouping dendrogram dendrogram ((B)) of triangle D3-A_PhPPh. D3-A_PhPPh.

The D4-A_CaPCa triangle was the only one that did not reject the null hypothesis, with an associated p-value of 0.1147. In the radial graph of Figure 10A, the discrimination between both types of mushrooms was not clear. This was confirmed by the detection of more than two groups by the hcluster function, which evidenced the impossibility of differentiating one mushroom from another Foods 2019, 8, 414 16 of 20 by grouping, as observed in Figure 10B. In Table6, where the aroma compounds of mushrooms are indicated, it can be seen that A. caesarea (Ca) and A. pantherina (P) had the closest profile. This may suggest that there are similarities between the volatile organic compounds that emerge from them. However,Foods 2019, 8 in, 414 contrast, in the corresponding sensory trial, a high level of discrimination was obtained.16 of 19

(A) (B)

FigureFigure 10. 10.D4-A_CaPCa: D4-A_CaPCa: thethe onlyonly triangletriangle thatthat diddid notnot rejectreject nullnull hypothesishypothesis inin thethe hierarchicalhierarchical clustercluster analysisanalysis (HCA) (HCA) analysis. analysis. ((AA)) Radial Radial graph; graph; ( B(B)) dendrogram. dendrogram.

3.4. ResultsDespite Comparison this particular situation, it could be statistically affirmed that the electronic nose was able to differentiate between the vast majority of the analyzed mushrooms with enough clarity. This In the “% Difference” column of Table 8, the p-value corresponding to the electronic nose was answers affirmatively the second question presented: machines can help us to find odorous differences subtracted from the p-value corresponding to the sensory test. The six positive results indicate an in mushrooms. equal or better performance of human olfaction in olfactory differentiation of these triangles, while 3.4.the Resultsnegative Comparison results indicate the opposite. It can be observed that human smelling discriminated with a confidence level of over 99% in 8 triangles, while the machine achieved 99.9% in 10 triangles. These In the “% Difference” column of Table8, the p-value corresponding to the electronic nose was odorous discrepancies between humans and machines, which are shown in Figure 11, suggest that subtracted from the p-value corresponding to the sensory test. The six positive results indicate an they move in a slightly different odorous spectrum. Figure 11 also reveals an improvement in the equal or better performance of human olfaction in olfactory differentiation of these triangles, while the results of sensory analysis over time, as previously outlined. negative results indicate the opposite. It can be observed that human smelling discriminated with a confidence level of over 99% in 8 triangles, while the machine achieved 99.9% in 10 triangles. These odorous discrepancies between humans and machines, which are shown in Figure 11, suggest that they move in a slightly different odorous spectrum. Figure 11 also reveals an improvement in the results of sensory analysis over time, as previously outlined.

Foods 2019, 8, 414 17 of 20

FigureFigure 11. 11. Comparison of of the the results results of the of triangular the triangular sensory sensory test to those test obtained to those by obtained the electronic by the electronicnose. nose.

4. Conclusions From a statistical point of view, the capacity of both human and electronic devices to discriminate between the species of wild mushrooms analyzed in this study was demonstrated. Now, although the study fulfilled the proposed objective, it was limited to a small number of species. In addition, they were of a single genus, were localized temporally and geographically, and were discriminated against (but not recognized). This should be the next working approach. The learning detected in the discrimination of mushrooms through the sense of human smell at the end of the sensory analysis supports studies that suggest its potential, although this has already been demonstrated in other fields, such as oenology, where an interesting wheel of aromas has been developed [48]. There is a need to work on the development of a similar tool in the field of mycology, which would provide olfactory education aimed at (together with other markers) the recognition of species of mushrooms, especially toxic ones. A limitation in this aspect is the lack of specialized tasters in mushrooms, which are needed to progress in sensory evaluations with a more selective approach. With regard to the development of electronic noses, support for the continuous improvement of different technologies of gas sensors and continuous research on possible applications could allow for the production of devices that recognize mushrooms by smell, for example, the edible Amanita ponderosa, which is very characteristic for its earthy smell and is sometimes confused with the Amanita verna, which is odorless and very toxic. The risks involved in this type of classification mean that the methods used must be totally contrasted with and complemented by other mushroom recognition tools. A further future application could be to characterize the state of ripeness of edible mushrooms. 1

Foods 2019, 8, 414 18 of 20

Other issues in which advances are required are the collection of sensor data and their speciﬁcity for the analysis of volatile compounds present in mushrooms, where there is a lack of information in the corresponding datasheets. The preprocessing technique applied in this study (fractional baseline manipulation) does not consider several features of the signal obtained directly from the sensor, and they may provide relevant data for the discrimination of all types of volatiles. To conclude, it should be noted that, as emerging research and new applications have revealed, electronic noses, from which valuable information will be obtained in a wide range of environments, will not take long to become an integrated element in our everyday devices. In addition, it is highly desirable that they could be used to promote the development of the smell sense of humans by supplying a feedback line to improve responses to diﬀerent odorous sensations. As a result, it would be possible to increase olfactory capacity and therefore human knowledge.

Author Contributions: F.P.-C.: investigation and formal analysis; P.A.: writing—original draft preparation and validation; J.I.S and J.L.: supervision, writing—review and editing. Funding: This research received no external funding. Acknowledgments: The authors want to thank the Centro de Profesores de Badajoz, the Sociedad Micológica Extremeña, and collaborators in the triangular sensory test. F. Portalo also wants to thank Meña. Conﬂicts of Interest: The authors declare no conﬂicts of interest.

References

1. Niimura, Y. Olfactory Receptor Multigene Family in Vertebrates: From the Viewpoint of Evolutionary Genomics. Curr. Genom. 2012, 13, 103–114. [CrossRef][PubMed] 2. McGann, J.P. Poor human olfaction is a 19th-century myth. Science 2017, 356, 7263. [CrossRef][PubMed] 3. Bushdid, C.; Magnasco, M.O.; Vosshall, L.B.; Keller, A. Humans can Discriminate more than one Trillion Olfactory Stimuli. Science 2014, 343, 1370–1372. [CrossRef][PubMed] 4. Gross, M. Our sense of smell at the crossroads. Curr. Boil. 2015, 25, R173–R176. [CrossRef][PubMed] 5. Majid, A.; Burenhult, N. Odors are expressible in language, as long as you speak the right language. Cognition 2014, 130, 266–270. [CrossRef][PubMed] 6. Jong, S.C.; Birmingham, J.M. Mushrooms as a source of natural flavor and aroma compounds. Mushroom Biol. Mushroom Prod. 1993, 345, 366. 7. Willis, C.M.; Britton, L.E.; Harris, R.; Wallace, J.; Guest, C.M. Volatile organic compounds as biomarkers of bladder cancer: Sensitivity and specificity using trained sniffer dogs. Cancer Biomarkers 2011, 8, 145–153. [CrossRef][PubMed] 8. Pardo, L.F.M.; Acosta, A.P. Detección de explosivos con la ayuda de animales: una revisión de la literatura científica. Rev. Logos Cienc. Tecnol. 2009, 1, 107–117. 9. García, A.; Sánchez, J.A. Setas de la Península Ibérica y de Europa; Everest: León, Spain, 2009; p. 402. 10. Romero de la Osa, L. La setas del Parque Natural Sierra de Aracena y Picos de Aroche; Diputación Huelva: Huelva, Spain, 2003; ISBN 978-84-8163-340-5. 11. Palazón, F. Setas para todos. Pirineos. Península Ibérica; Pirineo: Huesca, Spain, 2001; pp. 345–347. 12. Tulloss, R.E. Amaniteae: Amanita, Limacella, & Torrendia. Mycotaxon. 2005, 92, 469–496. 13. Eren, S.H.; Demirel, Y.; Ugurlu, S.; Korkmaz, I.; Aktas, C.; Güven, F.M.K. Mushroom poisoning: retrospective analysis of 294 cases. Clin. 2010, 65, 491–496. [CrossRef] 14. Moreno Raimundo, M.D. Instituto Nacional de Toxicología y Ciencias Forenses 2016. Ministerio de Justicia Secretaría General Técnica; 2017. Available online: https://www.mjusticia.gob.es/cs/ Satellite/Portal/1292428739049?blobheader=application%2Fpdf&blobheadername1=Content-Disposition& blobheadername2=Grupo&blobheadervalue1=attachment%3B+filename%3DMemoria_INTCF_2016. PDF&blobheadervalue2=INTCF (accessed on 12 September 2019). 15. Piqueras-Carrasco, J. Intoxicaciones por setas, una actualización. Revista Española de Medicina Legal 2014, 40, 19–29. [CrossRef] 16. Anzaldúa, M.A. La Evaluación Sensorial de los Alimentos en la Teoría y la Práctica; Acribia: Zaragoza, Spain, 1994; p. 198. Foods 2019, 8, 414 19 of 20

17. Hiraide, M.; Miyazaki, Y.; Shibata, Y. The smell and odorous components of dried shiitake mushroom, Lentinula edodes I: relationship between sensory evaluations and amounts of odorous components. J. Wood Sci. 2004, 50, 358–364. [CrossRef] 18. Xu, X.; Xu, R.; Jia, Q.; Feng, T.; Huang, Q.; Ho, C.-T.; Song, S. Identification of dihydro-β-ionone as a key aroma compound in addition to C8 ketones and alcohols in Volvariella volvacea mushroom. Food Chem. 2019, 293, 333–339. [CrossRef][PubMed] 19. Larsen, T.O.; Frisvad, J.C. Chemosystematics of Penicillium based on profiles of volatile metabolites. Mycol. Res. 1995, 99, 1167–1174. [CrossRef] 20. Malheiro, R.; De Pinho, P.G.; Soares, S.; Ferreira, A.C.D.S.; Baptista, P.Volatile biomarkers for wild mushrooms species discrimination. Food Res. Int. 2013, 54, 186–194. [CrossRef] 21. Jung, M.Y.; Lee, D.E.; Cheng, H.Y.; Chung, I.-M.; Kim, S.-H.; Han, J.-G.; Kong, W.-S. Characterization of Volatile Profiles of Six Popular Edible Mushrooms Using Headspace-Solid-Phase Microextraction Coupled with Gas Chromatography Combined with Chemometric Analysis. J. Food Sci. 2019, 84, 421–429. [CrossRef] 22. Zhou, J.; Feng, T.; Ye, R. Differentiation of Eight Commercial Mushrooms by Electronic Nose and Gas Chromatography-Mass Spectrometry. J. Sensors 2015, 2015, 1–14. [CrossRef] 23. Bhandari, M.P.; Carmona, E.N.; Abbatangelo, M.; Sberveglieri, V.; Duina, G.; Malla, R.; Comini, E.; Sberveglieri, G. Discrimination of Quality and Geographical Origin of Extra Virgin Olive Oil by S3 Device with Metal Oxides Gas Sensors. Proc. 2018, 2, 1061. [CrossRef] 24. Arroyo, P.; Herrero, J.L.; Suárez, J.I.; Lozano, J. Wireless Sensor Network Combined with Cloud Computing for Air Quality Monitoring. Sensors 2019, 19, 691. [CrossRef] 25. Pearce, T.C.; Schiffman, S.S.; Nagle, H.T.; Gardner, J.W. Handbook of machine olfaction: electronic nose technology; John Wiley & Sons: Hoboken, NJ, USA, 2002; ISBN 9783527303588. 26. G˛ebicki,J.; Szulczyński,B. Discrimination of selected fungi species based on their odour profile using prototypes of electronic nose instruments. Measurement 2018, 116, 307–313. [CrossRef] 27. Pei, F.; Yang, W.; Ma, N.; Fang, Y.; Zhao, L.; An, X.; Xin, Z.; Hu, Q. Effect of the two drying approaches on the volatile profiles of button mushroom (Agaricus bisporus) by headspace GC–MS and electronic nose. LWT 2016, 72, 343–350. [CrossRef] 28. Suarez, J.I.; Arroyo, P.; Lozano, J.; Herrero, J.L.; Padilla, M. Bluetooth gas sensing module combined with smartphones for air quality monitoring. Chemosphere 2018, 205, 618–626. [CrossRef][PubMed] 29. Herrero, J.L.; Lozano, J.; Santos, J.P.; Suárez, J.I. On-line classification of pollutants in water using wireless portable electronic noses. Chemosphere 2016, 152, 107–116. [CrossRef] 30. Instituto Nacional de Toxicología y Ciencias Forenses. Inquiry about mushroom manipulation intoxications; Noref M/E16/18356; Ministerio de Justicia: Madrid, Spain, 2016. 31. Bernhard, R.A.; Simone, M.J. THE LOCUS OF AROMA IN THE MUSHROOM: (Agaricus campestris L.). J. Food Sci. 1959, 24, 165–166. [CrossRef] 32. Kramida, A.; Ralchenko, Y.; Reader, J.; NIST ASD Team. Atomic Spectra Database (version 5.6.1). Available online: https://www.nist.gov/pml/atomic-spectra-database (accessed on 21 July 2009). 33. Olivas-Gastélum, R.; Nevárez-Moorillón, G.V.; Gastélum-Franco, M.G. Las pruebas de diferencia en el análisis sensorial de los alimentos. Tecnociencia Chihuahua 2009, 3, 1–7. 34. Mielle, P.; Marquis, F. An alternative way to improve the sensitivity of electronic olfactometers. Sensors Actuators B: Chem. 1999, 58, 526–535. [CrossRef] 35. Gutierrez-Osuna, R. Pattern analysis for machine olfaction: a review. IEEE Sensors J. 2002, 2, 189–202. [CrossRef] 36. Leenen, I. La prueba de la hipótesis nula y sus alternativas: revisión de algunas críticas y su relevancia para las ciencias médicas. Investig. en Educ. médica 2012, 1, 225–234. 37. Roessler, E.B.; Warren, J.; Guymon, J.F. Significance in triangular taste tests. Food Res. 1948, 13, 503–505. [CrossRef] 38. Roessler, E.B.; Pangborn, R.M.; Sidel, J.L.; Stone, H. Expanded Statistical Tables for Estimating Significance in Paired? Preference, Paired? Difference, Duo? Trio and Triangle Tests. J. Food Sci. 1978, 43, 940–943. [CrossRef] 39. Ghasemi-Varnamkhasti, M.; Mohtasebi, S.S.; Rodriguez-Mendez, M.L.; Lozano, J.; Razavi, S.; Ahmadi, H. Potential application of electronic nose technology in brewery. Trends Food Sci. Technol. 2011, 22, 165–174. [CrossRef] Foods 2019, 8, 414 20 of 20

40. Mirzaei, A.; Leonardi, S.; Neri, G. Detection of hazardous volatile organic compounds (VOCs) by metal oxide nanostructures-based gas sensors: A review. Ceram. Int. 2016, 42, 15119–15141. [CrossRef] 41. Granato, D.; Santos, J.S.; Escher, G.B.; Ferreira, B.L.; Maggio, R.M. Use of principal component analysis (PCA) and hierarchical cluster analysis (HCA) for multivariate association between bioactive compounds and functional properties in foods: A critical perspective. Trends Food Sci. Technol. 2018, 72, 83–90. [CrossRef] 42. Müllner, D. Modern hierarchical, agglomerative clustering algorithms. arXiv 2011, arXiv:1109.2378. 43. Suzuki, R.; Shimodaira, H. Pvclust: An R package for assessing the uncertainty in hierarchical clustering. Bioinformatics 2006, 22, 1540–1542. [CrossRef] 44. Sokal, R.R.; Rohlf, F.J. The Comparison of Dendrograms by Objective Methods. Taxon 1962, 11, 33–40. [CrossRef] 45. Santana, O.F. El Análisis de Cluster: aplicación, interpretación y validación. Pap. Rev. de Sociol. 1991, 37, 65. 46. Maga, J.A. Mushroom flavor. J. Agric. Food Chem. 1981, 29, 1–4. [CrossRef] 47. Macrae, A. Confidence intervals for the triangle test can give reassurance that products are similar. Food Qual. Prefer. 1995, 6, 61–67. [CrossRef] 48. Noble, A.C.; Arnold, R.A.; Buechsenstein, J.; Leach, E.J.; Schmidt, J.O.; Stern, P.M. Modification of a standardized system of wine aroma terminology. Am. J. Enol. Vitic. 1987, 38, 143–146.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).