Nondestructive measurement of squalene in olive oil by near infrared spectroscopy

1 Nondes uc i e measu emen o squalene in oli e oil by nea in a ed 1 spec oscopy 2 José A. Cayuela1, Juan F. Ga cía2 3 1Ins i u o de la G asa, CSIC 4 Campus o he Uni e si y ‘Pablo de Ola ide’, Ed. 46. 41013 Se ille, Spain 5 2Depa men o Chemical Enginee ing, Uni e si y o Se ille, C/ P o eso Ga cía González, 1, 41012 6 Se ille, Spain 7 1Co esponding au ho : [email protected] 8 ABSTRACT 9 This s udy se s he basis o de eloping a apid echnique o measu ing oli e oil 10 squalene, which is a heal hy compound. This echnique, based on nea in a ed 11 spec oscopy, is en i onmen ally iendly. The mos sui able wa eleng h anges we e 12 de ined, s udying he possible con ibu ion om he isible spec a. Fo his pu pose, 13 Pa ial Leas Squa es analysis was independen ly se up using wo op ical a angemen s, 14 wi h wa eleng hs 350-2500 nm and 1100-2300 nm. Models om only nea in a ed 15 wa eleng hs ga e he bes ou comes. The ex e nal alida ion exe cise o es ima ing 16 oli e oil squalene was sa is ac o y, wi h 2 0.83 and esidual p edic i e de ia ion 2.31. 17 The esul s sugges he p oposed echnique is use ul o es ima ing oli e oil squalene 18 con en . A so ing es o oli e oil in wo classes acco ding o i s squalene con en was 19 ca ied ou , wi h h eshold in 5.0 g.kg-1, using he model buil . The success o his 20 classi ica ion was 90%. 21 Keywo ds: classi ica ion; en i onmen iendly; NIR; PLS model; ; h eshold. 22 23 1. In oduc ion 24 Squalene (Figu e 1) is a i e pene alipha ic hyd oca bon, and i was named because o 25 i s p o usion in sha k li e oil, i s iches sou ce, whe e i eaches 900 g·kg-1. Sha k li e 26 oil has long been used as a adi ional heal h ood in Japan, wi h a pa icula bene i o 27 ascula heal h (Hamada e e al., 2015). Squalene is widely dis ibu ed in na u e, 28 especially in ege able oils such as oli e oil, palm oil, whea -ge m oil, ama an h oil, o 29 ice b an oil (Huang, Lin, & Fang, 2009). The e o e, by using di e en ex ac ion 30 Pos p in o LWT - Food Science and Technology Volume 88, Feb ua y 2018, Pages 103-108 DOI: 10.1016/j.lw .2017.09.047 2 me hods, ege ables o ma ine animals can be sui able squalene sou ces (Vázquez, 31 To es, Fo na i, Seno ans, & Regle o, 2007). 32 The main pa o i gin oli e oil is he saponi iable ac ion, a lipid ma ix o 33 iglyce ides, diglyce ides and monoglyce ides accoun ing o 985-995 g·kg-1 34 (Ci an os, 1999). Squalene is in ela i ely high quan i ies wi hin he oli e oil mino 35 ac ion. Eisne , I e son, Mozingo, & Fi es one (1965) s a ed ha squalene makes up 36 a ound 85-90% o he hyd oca bon ac ion o oli e oils. Besides, i makes up 60–75% 37 o he oli e oil unsaponi iable ac ion, in concen a ions be ween 0.2 and 7.5 g·kg-1 38 (Tisco nia, & E angelis i, 1982). 39 Figu e 1 40 One o he mos impo an di e ences be ween oli e oil and plan seed oils is squalene. 41 Compa ed o seed oils, oli e oil is an impo an sou ce o squalene. In o he edible 42 ege able oils, squalene makes up only 0.02–0.3 g·kg-1 (Rao, Newma k, & Reddy, 43 1998). Thus, oli e oil con ains 7 o 300 old mo e squalene han o he ege able oils and 44 up o 5000 old mo e han some ege able oods (Liu, Ah ens, Sch eibman, & C ouse, 45 1976). The e o e, i gin oli e oil may be a pa o he human die especially ich in 46 squalene. 47 Besides, he squalene con en a ies widely depending on he oli e oil p oduc wi h a 48 ange o 2 o 7 g·kg-1 (Rao, Newma k, & Reddy, 1998). A signi ican di e ence 49 be ween he ex a i gin class (EVOO) and he i gin class (VOO) has been epo ed, 50 wi h he la e ha ing mo e squalene han he e ined oli e oils (Owen, Mie , Giacosa, 51 Hull, Spiegelhalde , & Ba sh, 2000). Ne giz & Çelikkale (2011) showed ha e ining 52 educes he squalene con en . Fu he mo e, hey poin ed ou ha he majo dec ease in 53 squalene in ege able oils wi hin he e ining s eps occu s du ing oils’ deodo iza ion. 54 Oli e g owing echniques (Psomiadou & Tsimidou, 1999), oli e ui a ie y (Ne giz, & 55 Ünal, 1990) and ex ac ion (Ne giz, & Ünal, 1990; Samaniego-Sánchez, Quesada-56 G anados, López-Ga cía de la Se ana, & López-Ma ínez, 2010) in luence he le el o 57 squalene. Squalene ac s as a weak an ioxidan in oli e oil (Owen, Mie , Giacosa, Hull, 58 Spiegelhalde , & Ba sh, 2000). Thus, Psomiadou and Tsimidou (1999) p oposed ha 59 squalene con ibu es o oli e oil s abili y in a small quan i y, e en a low empe a u es. 60 3 The e is mul iple scien i ic e idence on he bene icial e ec s ha he in ake o squalene 61 om ood has on heal h (Newma k, 1997; Lasekan, Clay on, Gend on, & Ney, 1990; 62 Smi h, 2000; Os lund, Race e, & S enson, 2002; S andbe g, Til is, & Mie inen, 1990; 63 Smi h, 2000; He, & Co ke, 2003). Howe e , his ea u e o oli e oil has ecei ed li le 64 a en ion in he ma ke so a , since mos consume s a e unawa e on his ac . 65 The oli e oil indus y has g ea in e es on de e mining he quali y o oli e oil, using 66 as and eliable echniques. Besides, de eloping non-des uc i e echniques o educe 67 he expense o sol en s and eagen s is inc easingly impo an in an in e na ional 68 con ex o con e gence owa ds en i onmen al sus ainabili y. Among he a ious non-69 des uc i e solu ions o hese needs, nea -in a ed spec oscopy (NIRS) has made majo 70 achie emen s. NIRS is based on mul i a ia e models in which he spec al da a co ela e 71 wi h he analyzed ea u e. I p o ides se e al impo an ad an ages, as NIRS needs no 72 sol en s o eagen s, hus a oiding a majo expense, while being en i onmen ally 73 iendly. Addi ionally, NIRS is a apid, non-des uc i e, and po en ially mul i-pa ame e 74 me hod. 75 Se e al a icles on he use o NIRS and chemome ics o he analysis o di e en oli e 76 oil ea u es ha e been published in he ecen yea s (Nenadis & Tsimidou, 2017). S and 77 ou s udies di ec ed o cha ac e izing in ac oli es and oli e pas e o op imizing he 78 milling p ocess (Gio enzana e al., 2017), o con ol he quali y o oli e pomace oil 79 blended wi h palm oil used o deep- ying (Ben Hammouda, Z ibi, Ben Mansou , 80 Ma haus, & Bouaziz, 2017), as well as o he au hen ica ion and de ec ion o aud 81 (Ka una hilaka, Kia, S igley, Chung, & Mossoba, 2016). So ing oli e oil based on 82 alpha- ocophe ol and o al ocophe ol con en using NIRS has been ecen ly epo ed 83 (Cayuela & Ga cía, 2017). The NIRS abili y o analyze he majo oli e oil quali y 84 ea u es has been he subjec o se e al s udies (A men a, Ga igues, & De la Gua dia, 85 2007; Bendini, Ce e ani, Di Vi gilio, Belloni, Le cke , & Gallina-Toschi, 2007; Con e, 86 B ussolo, Pizzale, Ca azzolo, Meu ens, & Pa an, 2003; Cos a, Coelho, Gamba a, 87 Beze a, Ha op, & Ugulino, 2008; Cayuela, Mo eda & Ga cía, 2013). In ac , NIRS 88 echniques a e me hods o hese ou ine analyses in a g owing numbe o labo a o ies. 89 Howe e , he possibili y o measu ing squalene in oli e oil by NIRS has ne e been 90 epo ed up o da e. Besides, squalene NIR abso p ion bands ha e no ye been 91 desc ibed, o he bes o ou knowledge. 92 4 Since he concen a ion o squalene a ies widely among di e en oli e oils, he e is an 93 in e es in he de elopmen o apid echniques o dis inguish oli e oils acco ding o i s 94 con en . In ac , he indus y migh ha e an in e es in sepa a ing oli e oils acco ding o 95 di e en squalene con en s. The adi ional me hod o he analysis o squalene in oli e 96 oil is GC. Howe e , i is no usually pe o med in he oli e oil indus y, since squalene 97 is no conside ed in he egula ion o cha ac e ize he quali y o pu i y o oli e oil 98 (Eu opean Commission, 1991). The e o e, he e is a challenge on cha ac e izing oli e 99 oil ega ding squalene. This wo k se s up he basis o de eloping new apid NIRS 100 echniques o measu ing oli e oil squalene con en . I was con enien o cla i y i he e 101 a e any egions om he oli e oil’s isible spec um con ibu ing o model pe o mance, 102 since pu e squalene is a pale yellow liquid. The wa eleng hs ha con ibu e o 103 p edic i e models ha e been de ined. 104 2. Ma e ial and Me hods 105 2.1. Oli e Oils 106 A o al se o 180 oli e oil samples was made up om di e en o igins. High quali y 107 Ex a Vi gin Oli e Oils (EVOO) we e bough a oli e oil specialized shops; his g oup 108 con ibu ed wi h 32 samples, o which 27 we e a ie al and he emaining 5 we e 109 mix u es om di e en a ie ies. These EVOO we e used o elabo a e 17 addi ional 110 coupage samples. Oli e oils no mally ound in he ma ke we e also used; his g oup 111 was composed o 10 EVOOs, 40 Cu en Oli e Oils and 25 Pomace Oli e Oils. Oli e 112 oil samples we e p o ided also om a collabo a o indus y, con ibu ing wi h 14 113 EVOOs, 25 Vi gin Oli e Oils and 14 Lampan e Oli e Oils. The cha ac e is ics o he 114 oli e oil samples a e shown in Table S1. 115 2.2. Re e ence Analysis 116 Squalene analysis we e ca ied ou by Gas Ch oma og aphy (GC) acco ding o Lanzón, 117 Albi, Ce , & G acián (1994), modi ied acco ding o Mo eda, Pé ez-Camino, & Ce 118 (2004), a he Ins i u o de la G asa (CSIC). B ie ly, 0.1 mg o oli e oil sample was 119 disposed in a 4 mL sc ew ial, adding 1 mL o squalane 5 mg·mL-1 as he in e nal 120 pa e n. This was dissol ed in hep ane o comple e a olume o 3 mL and shaked gen ly 121 by hand. Then 200 L o me hanolic 2 mol·L-1 KOH was added, sepa a ing he aqueous 122 and lipid phases. The uppe phase was collec ed in o a 2 mL ch oma og aphy ial and 123 5 hen injec ed in o he GC ins umen . A GC HP-5890 (Hewle Packa d En e p ise, Palo 124 Al o, USA) equipped wi h a spli /spli less injec ion sys em was used wi h a SP-5 125 capilla y column 5% phenylme hylsilicone used silica, 30 m long, 0.25 mm in e nal 126 diame e and 0.25 m phase hickness, (Me ck, Da ms ad , Ge many). Flame ioniza ion 127 de ec o (FID) and so wa e Chem S a ion o he eco ding and p ocessing o da a we e 128 used. The analyses we e conduc ed wi h wo eplica es. The esul s we e gi en wi h one 129 signi ican digi . 130 2.3. Nea in a- ed spec oscopy 131 Op ical a angemen s NIRS and VIS/NIRS we e used o de ining he wa eleng hs 132 con ibu ing o he p edic i e models, especially o cla i ying he con ibu ion om 133 isible spec a. Besides, using wo di e en ins umen s allowed checking hei esul s, 134 beyond hei compa ison. 135 The samples’ spec a we e eco ded di ec ly om oli e oils wi hou any o he ea men . 136 The empe a u e o a body has an impo an in luence on NIR adia ion. The e o e, he 137 samples we e aken om 4 °C s o age and placed a oom empe a u e in he labo a o y 138 18h be o e p ocessing. A he mos a ic ba h ixed a 33 °C o 30 min held he 20 mL 139 sample con aine s o ensu e empe a u e s abili y. The a e aged spec um om wo 140 measu emen s o 50 spec a each was egis e ed, wi h each sample. The same p ocedu e 141 was used wi h bo h op ical con igu a ions. 142 Fo NIRS, he measu ing mode was pos dispe si e ans lec ance. A Lumina 143 (B im ose Inc., Ma yland, USA) spec ome e was used. This ins umen consis s o an 144 acous o-op ic unable il e (AOTF) wi h InGaAs de ec o (1100-2300 nm). The 145 e e ence is au oma ically aken, he scanning speed is 60 ms. The spec ome e is 146 composed o a hand-held uni , equipped wi h a base o labo a o y use. A ans lec ance 147 p obe accesso y was used. The p obe is in s ainless s eel, wi h h eaded in e changeable 148 op ical pa h. The spec a we e egis e ed as a whole, he spec al a iables ma ching a 2 149 nm in e als. The epea abili y o he ins umen , exp essed as he s anda d de ia ion o 150 he a e age abso bance o i e measu emen s o a whi e ile, is 6.76 10-4. The signals 151 we e cap u ed using Acqui e so wa e (B im ose Co p., Ma yland). 152 The VIS/NIRS was ca ied ou using a Labspec (Analy ical Spec al De ices Inc., 153 Boulde , USA) spec ome e , wi h ansmi ance op ical mode consis ing o a liquid 154 6 accesso y (Ocean Op ics, La go, USA). A qua z spec opho ome ic cu e e (Hellma 155 Analy ics, Müllheim, Ge many) wi h 10 mm pa h leng h held he samples. The whole 156 VIS/NIRS spec um (350−2500 nm) was egis e ed, wi h each spec al a iable 157 co esponding o a 2 nm in e al. The con igu a ion o 50 spec a in con inuous 158 acquisi ion was used. Indico P o so wa e (Analy ical Spec al De ices Inc., Boulde , 159 USA) was used o his pu pose. The spec ome e was equipped wi h h ee de ec o s. 160 The de ec o o he isible ange (350-1000 nm) was a ixed e lec i e holog aphic 161 diode a ay wi h a sensi i i y o 512 pixels. A holog aphic as scanne InGaAs de ec o 162 cooled a -25ºC co e ed he wa eleng h ange o 1000-1800 nm. The same de ice 163 coupled wi h a high o de blocking il e was used o he in e al 1800-2500 nm. The 164 scanning speed was 100 ms, and he acquisi ion p ocessing ime is less han a minu e 165 o each sample, all s eps included. The epea abili y, exp essed as he s anda d 166 de ia ion o he a e age abso bance o i e measu emen s o a whi e ile be ween 350 167 and 2500 nm, is 6.00 10-4. 168 2.4. Squalene spec um 169 Squalene o 98% pu i y (Me ck, Da ms ad , Ge many) was used as pa e n o 170 cha ac e ize he squalene VIS/NIR abso p ion bands. The pa e n spec um was 171 egis e ed only once using he Labspec spec ome e , by a e aging ou eplica es. The 172 es o he p ocedu e was he same as p e iously desc ibed o oli e oil samples. 173 2.5. Chemome y 174 Possible oli e oil g oups we e analyzed by P incipal Componen Analysis (PCA), 175 which was also used o de ec ing possible spec al ou lie s. I was ca ied ou om he 176 oli e oils spec a o bo h op ical con igu a ions using The Unsc umble 9.7 (CAMO 177 So wa e AS, Oslo, No way). 178 Mul i a ia e Pa ial Leas Squa es (PLS) analysis was pe o med om he spec al 179 a iables o nea in a ed (NIRS) and isible-nea -in a ed (VIS/NIRS), using he 180 squalene e e ence analysis as a dependen a iable. T ansmi ance spec al da a we e 181 a e aged o 8 nm in e als and ans o med in o abso bance, hen, mean no maliza ion 182 (MN), s anda d no mal a ia e no maliza ion (SNV), and i s (D1SG) and second 183 (D2SG) Sa i zsky−Golay de i a i es ea men s we e ca ied ou . The PLS models o 184 squalene we e buil om he a e aged and ea ed spec um using The Unsc umble 9.7. 185 7 The ull c oss in e nal alida ion (FCV) p ocedu e was used. The ou lie s we e 186 iden i ied as samples showing signi ican high esiduals, acco ding o The Unsc umble 187 9.7. Sco es plo we e displayed om he eg ession o e iew plo , hen selec ing he 188 wa ning lis op ion. The ou lie lis was displayed by clicking he ou lie s bu on. The 189 esiduals a e he di e ences be ween he p edic ed and he analyzed alues. 190 Two independen mul i a ia e calib a ion models o squalene p edic ion (M1 and M2) 191 we e es ablished om he spec a ob ained wi h NIRS and VIS/NIRS. The models’ 192 p incipal componen s (PCs) we e ixed a e he es s using 10 PCs a i s . 193 The calib a ion se o PLS models excluded he ex e nal alida ion se . I was de ined 194 as one hi d om he 180 oli e oil samples a ailable, coun ing om he i s . Six y 195 samples we e aken by including one o each h ee, om he da a base o The 196 Unsc amble 9.7., in he same o de as he samples we e included. This alida ion se 197 was andomly o med, since each sample was andomly egis e ed om a comple ely 198 independen oli e oil ba ch, e en when he mechanical selec ion by he so wa e i was 199 no andom. The squalene concen a ions ange in he alida ion se we e simila o ha 200 ange o he calib a ion, as i is shown o wa d. Those wa eleng hs whose co ela ion 201 wi h squalene con en was close o ze o we e emo ed in successi e PLS cycles, using 202 The Unsc amble 9.7. The a iable selec ion ended in he las cycle ha imp o ed he 203 squa ed coe icien o c oss alida ion o he calib a ion (R2CV). This p ocedu e 204 p o ided he spec al a iables selec ed o he PLS models. To assess model’s i ness, 205 he s anda d e o o calib a ion (SEC) and he closeness be ween hei squa ed 206 coe icien o calib a ion (R2) and R2CV we e conside ed. 207 2.6. Model Pe o mance Assessmen 208 Calib a ion models we e assessed by ex e nal alida ion exe cises. Fo his pu pose, 209 squalene con en was p edic ed in a p e iously ese ed se o med by 60 oli e oil 210 samples which did no pa icipa e in he mul i a ia e models. The model pe o mance 211 was assessed acco ding o he 2 om he ex e nal alida ion exe cises, which 212 co esponds o he simple linea eg ession be ween he analyzed and p edic ed alues. 213 A he same ime, he esidual p edic i e de ia ion (RPD) om he ex e nal alida ion 214 exe cise was conside ed. The RPD was de ined (Fea n, 2002) as he a io be ween he  215 om he e e ence da a o he alida ion se and he s anda d e o o pe o mance 216 8 (SEP). Also, he sepa a e analysis o he calib a ions o NIRS and VIS/NIRS allowed 217 con i ming hei p edic i e abili y. 218 2.7. Classi ica ion Tes s 219 Classi ica ion es s o oli e oil acco ding o hei squalene con en we e conduc ed by 220 using he PLS model as a quali a i e disc imina ion echnique. The spec al da a om 221 only he con igu a ion p o iding he bes yields we e used o his pu pose. Two classes 222 o oli e oils, High Squalene (HS), wi h squalene concen a ion abo e 5.0 g·kg-1, and he 223 o he one Low Squalene (LS), wi h squalene concen a ion below o equal o 5.0 g·kg-1, 224 we e ixed o classi ica ion es s in o wo squalene le els. These wo classes we e 225 de ined acco ding o he squalene mean o he o al sample se analyzed, whose alue 226 was his h eshold. The echnique pe o mance assessmen was by i s success g ade. 227 This was exp essed as he pe cen age o samples in which he p edic ed and ac ual 228 classi ica ions coincided. 229 3. Resul s and Discussion 230 3.1. Oli e Oil Spec um 231 Nea -in a ed spec a show a ious o e lapping bands, due o he i s and second 232 o e ones and a combina ion o he undamen al ib a ions, mainly ca bon–hyd ogen 233 (Shenk, Wo kman, & Wes e haus, 2001). Assigning he majo isible abso p ion bands 234 o oli e oil was done by Moyano, el ndez, Alba, & He edia (2008). Oli e oil spec a 235 om he samples analyzed in his wo k, shown in Figu e 2, a e consis en wi h he 236 p e iously indica ed epo s. A i s mino peak occu ed nea 415 nm. This a ea sui s 237 he wa eleng hs o oil abso p ion o da k blue colo ed ligh . I could be due mainly o 238 ca o enoids, as well as o pheophy in A, pheopho bide A and py opheophy in A. A 239 second peak was nea 450 nm, which co esponds o blue ligh abso p ion, a 240 cha ac e is ic o ca o enoids. A hi d peak appea ed a ound 670 nm, which coincides 241 wi h chlo ophylls abso p ion (Moyano, el ndez, Alba, & He edia, 2008). A b oad 242 abso bance band showed a ound 1220 nm, p obably due o second o e ones o C–H 243 and CH=CH– s e ching ib a ions o oil. A high in ensi y abso bance peak occu ed 244 a ound 2300 nm, caused by a combina ion o undamen al ib a ions o he C-H g oups 245 (Hou an , Bae en, Mo ales, Meu ens, & Apa icio, 2000). The squalene VIS/NIR spec a 246 egis e ed in his s udy a e shown in Figu e 3. The majo di e ences in he spec um o 247 9 squalene wi h espec o oli e oil co espond o he isible zone, since squalene is 248 almos colo less. On he con a y, in he NIR egion he peaks wi h squalene we e 249 p ac ically he same as in oli e oil, wi hou ema kable di e ences. As a hyd oca bon 250 molecule, a pa o he squalene C–H and CH=CH– s e ching ib a ions may be 251 o e lapping hose o iglyce ides, whose s uc u e is also ca bon–hyd ogen. Howe e , 252 he spa ial con igu a ion o he squalene molecule is clea ly di e en om ha o 253 iglyce ides due o i s speci ic bonds. Mo eo e , he la e ha e oxygen, while squalene 254 lacks his chemical elemen . The e o e, NIR abso p ion in ensi ies may e lec such 255 di e ences among di e en oli e oils, despi e he ac ha hey maybe no explici in 256 he spec um shape. 257 Figu e 2 –Figu e 3 258 3.2. Popula ion Cha ac e iza ion 259 The alues o he squalene e e ence analysis o he calib a ion and ex e nal alida ion 260 se s a e ga he ed in Table 1. As can be seen, a wide squalene a ia ion in eg a es in o 261 he calib a ions, anging om 1.01 g·kg-1 o 10.15 g·kg-1. The s a is ical mean o he 262 calib a ion se s analyzed was 5.02 g·kg-1 o squalene. The squalene con en in he oli e 263 oils analyzed in his wo k showed a ange e en wide han hose desc ibed in he 264 li e a u e (Tisco nia & E angelis i, 1982; Rao, Newma k & Reddy, 1998; Owen, Mie , 265 Giacosa, Hull, Spiegelhalde , & Ba sh, 2000). This ange e lec s he ex en o 266 a ia ion o squalene in his p oduc , which accoun s o he in e es o dis inguishing 267 oli e oils ha a e help ul o heal h due o hei high squalene con en . 268 Table 1 269 3.3. P incipal Componen Analysis 270 The oli e oils spec a PCA analysis o bo h spec ome e s showed he absence o 271 sample g oups. The PCA o NIRS, shown in Figu e 4, s ands ou o showing h ee 272 samples widely sepa a ed om he majo g oup. Two mo e samples appea ed sepa a ed 273 om he majo g oup, bo h consis en wi h he HS class. The emaining samples did no 274 show any consis en g ouping end. In he PCA o he VIS/NIRS spec a, shown in 275 Figu e 5, only wo samples appea clea ly sepa a ed. One o hese samples ma ches he 276 16 Tisco nia, E. F. M., & E angelis i, F. (1982). Chemical compsqualene ion o oli e oil 440 and i s a ia ions induced by e ining. Ri is a I aliana delle Sos anze G asse, 59, 519–441 56. 442 Rao, C. V., & Reddy, B. S. (1993). Modula ing e ec o amoun and ypes o die a y a 443 on o ni hine deca boxylase, y squalenene p o ein kinase and p os aglandins p oduc ion 444 du ing colon ca cinogenesis in male F344 a s. Ca cinognesis, 14, 1327–1333. 445 Vazquez, L., To es, C. F., Fo na i, T., Seno ans, F. J., & Regle o, G. (2007). Reco e y 446 o squalene om ege able oil sou ces using coun e cu en supe c i ical ca bon dioxide 447 ex ac ion. Jou nal o Supe c i ical Fluids, 40, 59–66. 448 449 Figu e cap ions 450 Figu e 1. Squalene s uc u e. 451 Figu e 2. Oli e oil isible and nea in a ed spec a om he samples analyzed. 452 Figu e 3. Visible and nea in a ed spec a o he squalene pe e n (pu i y 98.0%). 453 Figu e 4. P incipal Componen Analysis o he nea in a ed spec a (1100-2300 nm) 454 om he analyzed oli e oils. 455 Figu e 5. P incipal Componen Analysis o he isible and nea in a ed spec a (350-456 2500 nm) om he analyzed oli e oils. 457 Figu e 6. Spec al a iables con ibu ing o he squalene nea in a ed model (M1). 458 Figu e 7. Pa ial Leas Squa es quan i a i e squalene model (M1). 459 Figu e 8. P edic ion exe cise o he oli e oil squalene con en (V1). 460 461 Figu e 1 Figu e 2 Figu e 3 Figu e 4 Figu e 5 Figu e 6 Figu e 7 Figu e 8 Table 1 N  σ Range Calib a ion 118 5.10 2.19 1.01-10.15 Valida ion 59 4.88 2.33 1.22-10.02 Table 1. S a is ics o he oli e oil squalene con en (g·kg-1) o he calib a ion and ex e nal alida ion se s. N, sample se size; , mean; σ, s anda d de ia ion.