Delft University of Technology
Classification of condom lubricants in cyanoacrylate treated fingerprints by desorption
electrospray ionization mass spectrometry
van Helmond, Ward; Begieneman, Mark P.V.; Kniest, Roos; de Puit, Marcel
DOI
10.1016/j.forsciint.2019.110005
Publication date
2019
Document Version
Final published version
Published in
Forensic Science International
Citation (APA)
van Helmond, W., Begieneman, M. P. V., Kniest, R., & de Puit, M. (2019). Classification of condom
lubricants in cyanoacrylate treated fingerprints by desorption electrospray ionization mass spectrometry.
Forensic Science International, 305, [110005]. https://doi.org/10.1016/j.forsciint.2019.110005
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy
Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
Classi
fication
of
condom
lubricants
in
cyanoacrylate
treated
fingerprints
by
desorption
electrospray
ionization
mass
spectrometry
Ward
van
Helmond
a,b,c,*
,1,
Mark
P.V.
Begieneman
a,1,
Roos
Kniest
a,
Marcel
de
Puit
a,c,*
a
NetherlandsForensicInstitute,DigitalTechnologyandBiometrics,LaanvanYpenburg6,2497GB,DenHaag,theNetherlands
b
AmsterdamUniversityofAppliedSciences,ForensicScience,Weesperzijde190,1097DZ,Amsterdam,theNetherlands
c
DelftUniversityofTechnology,FacultyofAppliedSciences,DepartmentofChemicalEngineering,VanderMaasweg9,2629HZ,Delft,theNetherlands
ARTICLE INFO
Articlehistory: Received6August2019
Receivedinrevisedform16October2019 Accepted21October2019
Availableonline23October2019
Keywords:
MassSpectrometryImaging DESI-MSI
Polydimethylsiloxane Polyethyleneglycol Principalcomponentanalysis Lineardiscriminantanalysis
ABSTRACT
Tracesofcondomlubricantsinfingerprintscanbevaluableinformationincasesofsexualassault.Ideally, notonlyconfirmationofthepresenceofthecondombutalsodeterminationofthetypeofcondombrand usedcanberetrieved.Previousstudieshaveshowntobeabletoretrieveinformationaboutthecondom brandandtypefromfingerprintscontaininglubricantsusingvariousanalyticaltechniques.However,in practice fingerprintsoftenappearlatentand needtobe detected first,whichis oftenachievedby cyanoacrylate fuming. In this study, we developed a desorption electrospray ionization mass spectrometry (DESI-MS) method which, combined with principal component analysis and linear discriminantanalysis(PCA-LDA),allowsforhighaccuracyclassificationofcondombrandsandtypesfrom fingerprints containing condom lubricant traces. The developed method is compatible with cyanoacrylate(CA)fuming.Wecollectedandanalyzedarepresentativedataset fortheNetherlands comprising32differentcondoms.Distinctivelubricantcomponentssuchaspolyethyleneglycol(PEG), polydimethylsiloxane(PDMS),octoxynol-9andnonoxynol-9werereadilydetectedusingtheDESI-MS method.Basedontheanalysisoflubricantspots,a99.0%classificationaccuracywasachieved.When analyzinglubricantcontainingfingerprints,anoverallaccuracyof90.9%wasobtained.Fullchemical imagescouldbegeneratedfromfingerprints,showingthedistributionoflubricantcomponentssuchas PEGandPDMSthroughoutthefingerprint,whilestillallowingforclassification.Thedevelopedmethod showspotentialforthedevelopmentofDESI-MSbasedanalysesofCAtreatedexogenouscompounds fromfingerprintsforuseinforensicscience.
©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1.Introduction
Sexual assault is a major health problem and a violationof human rights [1]. When the identity of the perpetrator is unknown,themostimportanttracesin sexualassaultcasesare oftenofbiologicalorigin,suchasblood,semen,salivaandhair.This isduetothefactthatDNAcanberetrievedand analyzedfrom thesetraces,allowingfortheidentificationofpossiblecriminals. Nowadays,biologicalmaterialbecomesincreasinglyhardertofind incases ofsexual assault,potentiallybecauseawarenessof the importance ofDNA in forensictechniques and evidence in the justicesystem is growing among criminals [2]. Because of this
awareness, as wellas alertnessto sexual transmitted diseases, criminalsarebecomingmorevigilantinleavingbiologicaltraces andtheuseofcondomsinsexualassaultcaseshasincreasedover thepastdecades[3–6].Othertypesoftraceevidencemaythusbe requiredtoestablisha linkbetweenvictimandcriminal.Inthe past,studieshaveshownthatacondomcanbeacriticalpieceof evidenceinsexual assaultcases[4,7,8]. Theanalysisofcondom derived traces can thus be of significant associative evidential importance.Ideally,notonlyconfirmationofthepresenceofthe condombutalsodeterminationofthetypeofcondombrandused canberetrieved.
In thelast decades several studies have beenperformed to
develop methods to detect traces of condoms. Most of these
studiesfocusedonthelubricantsthatareaddedtothecondomsby manufacturers.Indeed,primarycomponentsoflubricantssuchas polydimethylsiloxane(PDMS)andpolyethyleneglycol(PEG)were foundto bedetectable bydesorption chemical ionizationmass spectrometry[9],pyrolysisgaschromatographymass spectrome-try(pyGC-MS),GC–MS[10],Ramanspectroscopy[11]andFourier
* Correspondingauthorsat:NetherlandsForensicInstitute,DigitalTechnology andBiometrics,LaanvanYpenburg6,2497GB,DenHaag,theNetherlands.
E-mailaddresses:w.van.helmond@hva.nl(W.vanHelmond),m.de.puit@nfi.nl
(M.dePuit).
1
Theseauthorscontributedequally.
http://dx.doi.org/10.1016/j.forsciint.2019.110005
0379-0738/©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
ForensicScienceInternational305(2019)110005
ContentslistsavailableatScienceDirect
Forensic
Science
International
transforminfrared spectroscopy (FT-IR) [9,12]. Another specific componentincaseofspermicidecontainingcondoms, nonoxynol-9(N9),couldalsobeidentifiedbyFT-IR[9],GC–MS[13]andliquid
chromatography mass spectrometry (LC–MS) [12]. Multiple
studiesexamined thepossibilitytodiscriminatedifferenttypes ofcondoms.Maynardetal.describedatwo-stepmethodusing FT-IRasafirstscreeningtool,followedbyeitherGC–MS,pyGC-MSor
LC–MS as a confirmation method, enabling them to uniquely
identify 11 types of condoms [12]. Burger et al. showed that capillaryelectrophoresesmayalsobeapromisingtechniquefor classifying both condom and personal lubricants, although it remainedunclearwhichdiscriminatingconstituentswereusedin theanalysis[14].
However,mostoftheseanalytical techniquesrequiresample preparationand/orextractionwhichmaybetimeconsumingand resultinlossoftheinitialtraceevidence.Additionally,themethod itselfmayalsolimittheamountofinformationretrievedfromthe sample. For instance, analysisof silicone lubricants by GC–MS requirespyrolysisofthelubricant[10],thatcancausedegradation ofminorcomponents.Theseminorcomponentswerefoundtobe an important differentiating factor in distinguishing sexual lubricantsandpersonalhygieneproducts,whichcontainsimilar majorcomponentssuchasPDMSandPEG[15].Also,preservation oftheoriginalevidentialtracecanbeofgreatinterestinforensic science. In this respect, the use of ambient ionization mass spectrometrytechniquesismorefavorable.Apopulartechnique thathasbeenusedinrecentyearsisdirectanalysisinrealtime (DART)MS,thathasbeenshowntobeaveryeffectivetoolforthe
detectionofboth themajorand minorcomponentsof condom
lubricants,withouttheneedtoextensivelypreparethesampleor potentiallossofevidence[16–20].Furthermore,DARTanalysisis highlyeffectiveindiscriminatinglubricants.Baumgartenetal.and Maric et al. successfully discriminated condom and personal lubricantsusing DART-Time-of-Flight (TOF) MSanalysis[19,18]. Using a DART-High Resolution MS (HR-MS) analysistechnique, Coonetal.couldrapidlygeneratediagnosticchemicalfingerprint
signatures of 110 condoms, enabling them to discriminate
condomsof16differentbrands[20].However,adisadvantageof DARTanalysisisthatthistechniqueisunabletoretrievespatial
chemical information from the samples, such as fingerprints
containinglubricants.
Lubricatedfingerprintsarelikelytobefoundatacrimesceneof sexualassault, ashandlingof a condomwill transfertheouter
coating of the condom onto the perpetrator’s fingerprints,
potentiallyleaving condomlubricant contaminated fingerprints behind[21,22].Detectionofalubricantfromafingerprintfoundat asexualassaultscene,wouldgreatlyincreasethestrengthofthe evidence,asitnotonlyestablishescontactwithacondombutalso indicatesthepresence of thecriminal atthe crimescene [21]. Bradshaw et al. developed a method for the visualization of condomlubricantwithinafingerprintbymappingthefingerprint ridgepattern using Matrix AssistedLaser Desorption/Ionization Mass Spectrometry Imaging (MALDI-MSI) [21]. In a follow-up studytheyshowedthatdifferentiationof6differentcondomsin
lubricated fingerprints was possible using MALDI-MSI, Raman
spectroscopyandATR-FTIRimaginginasynergisticmanner[22]. BesidesMALDI-MSI,desorptionelectrosprayionization(DESI)MSI hasalso beenshowntobea powerfulanalytical tool. Whereas MALDIoffersexcellentspatialresolution20
m
m,oneofthemain disadvantagesisthe necessitytoapplymatrix solutiontoyour sample[23].Applyingthematrixsolutionintoolargedropletswill bedetrimentalforthespatial resolutionas diffusionwill occur withinthesedroplets.OneofthemainadvantagesofDESI-MSIis thatsamplesrequire nosamplepreparation andcan bereadily analyzed.UsingDESI,typicallyspatialresolutionsof100m
mcan beachieved[23].However,thechoiceoftheelectrospraysolventcompositionisessential,astheinteractionbetweenthe electro-sprayandthesurfacegreatlyinfluencesthesensitivityandspatial resolution [24]. Mirabelliet al. wereable togeneratechemical images of latent lubricated fingerprints deposited on different surfaces and of different ages using DESI-MS [25]. The data
acquired from DESI-MS analysis of lubricated fingerprints,
combinedwithsupervisedpatternrecognitionstatisticalanalysis (lineardiscriminantanalysis(LDA)andsoftindependentmodeling ofclassanalogy(SIMCA)),enabledMirabellietal.todistinguish10 differentcondomtypeswitha94%predictionabilityforbothLDA andSIMCA[26].
However,fingerprintsinpracticeoftenneedtobedetectedfirst,
as they appear latent. One of the most used visualization
techniques for latent fingerprints on non-porous substrates is cyanoacrylate (CA) fuming [27,28]. Fingerprints from a sexual assault scene that have been analyzed at a forensic lab, are potentiallytreatedwithCA.Inthisregard,theaimofourstudywas todevelopamethodtoanalyzelubricatedfingerprints,whichwas compatible with CA fuming and able to differentiate between differenttypesofcondoms.Thedevelopedmethod,basedon DESI-MSanalysis,is capableofgeneratingchemical images,mapping
common lubricant components in a lubricated fingerprint.
Additionally, using this method, combined with a statistical approach, PCA followed by LDA, we were able to differentiate between32typesofcondomsfrom21differentbrands.
2.Materials&methods 2.1.Materials
UPLC-gradeacetonitrileandformicacidwerepurchasedfrom Biosolve(Valkenswaard,Netherlands).UPLC-grademethanolwas
purchased at Merck (Darmstadt, Germany). Cyanoacrylate was
purchasedfromBVDA(Haarlem, Netherlands).Microscopeglass slides were purchased from Thermo Fischer Scientific (Breda,
Netherlands). 24-wells slides were purchased from Prosolia
(Indianapolis, USA). Reference masses purine (5mM) and hex-akis(1H,1H, 3H-tetrafluoropropoxy)phosphazine)(HP-0921, 2.5 mM)werepurchasedfromAgilentTechnologies(SantaClara,USA). Mitramicrosamplertips(10
m
L)werepurchased fromNeoteryx (Torrance,USA).Arangeof32differentcondomswerepurchased atonlinepharmacyandcondomwebsites(Table1).2.2.Lubricantsamples
Lubricantofeachcondomwascollectedbyswipingtheinterior ofthecondompackageandbothsidesofthecondomwitha10
m
L microsampleruntilsaturation.Themicrosamplerisavolumetric absorptionmicrosamplingdevice(VAMS),thatonlyabsorbs10m
l of sample. Because of this, it provides more control over the amountoflubricanttobesampled,incomparisontocottonswabs,that absorbed too much lubricant. The condom lubricant was
carefullytransferredontothe24-wellsslidebyslightlytouching
each well once with the lubricated microsampler tip. Each
lubricantwas spotted 12times (n=12) onseparate wells. Asa control,eachsamplewellwasfollowedbyablankwell.Slideswere left to dry for at least 1h at room temperature (RT). Next, cyanoacrylate(CA,0.5gheatedto120C)fumingwasperformed onallslides,inaMVC1000fumingsystem(FosterandFreeman LTD,Worcestershire,UK)for10minat80%humidity.Slideswere thenlefttodryovernightatRTbeforeanalysis.
2.3.Fingerprintsamples
Fingerprintsweredonatedvoluntarilybya femaleandmale donor, after giving informed consent. No ethical approval was
obtainedasthematerialwasgatheredinanoninvasivemanner anddidnotinfringeonanyprivacyofthedonors.Allexperiments werecarriedoutfollowinginstitutionalguidelinesandaccording torelevantlaws.Allcondoms(n=32)werehandledbyeachdonor (n=2).Aftertouchingthecondomandtheinsideofthepackaging, thelubricantwasdistributedoverthefinger.After5minofdrying, fingerprintsofeachcondomlubricantweredepositedon micro-scopeslides.Fromeachdonor,ablankfingerprintwasusedasa control(n=2).Slideswerelefttodryforatleast1hatRTfollowed byCAfuming,asdescribedabove.Fingerprintswerethenleftto dryovernightatRT.
2.4.DESI-Q-TOFMS
Desorptionelectrosprayionizationmass spectrometry (DESI-MS)datawereacquiredusinganAgilenttechnologies(SantaClara, USA)6530quadrupoletime-of-flight(TOF) MSequippedwitha Prosolia(Indianapolis,USA)2D-DESI.24-wellssampleslideswere analyzedindwellmodeusingpositivepolaritywiththefollowing parameters:sprayvoltage,5kV;nitrogensheathgaspressure,6.0 bar;dryinggasflow,8L/minute,sourcegastemperature,300C; acquisition time, 200ms; mass range, m/z 100–1200; inlet-to-surface distance, 1mm and tip-to-surface distance, 3mm. Combinationsofseveralsprayincidentangles(52,45 and35) and tip-to-inletdistances(4, 5,6, and8mm)weretested.Best results were achieved with an angle of 45 and a tip-to-inlet distanceof6mm.Alargertip-inletdistanceledtodecreased carry-over,alsodescribedbyMirabellietal.[26].Adwelltimeof20swas used,withapost-acquire-delaytimeof30sinbetweenthewells. Tofurtheravoidcarry-overafteranalysisofeachsamplewell,the nextblankwellwasdwelledfor5minbeforemeasuringthenext sample.Additionally,theMS-inletwascleanedafteranalysisof3– 4slidestoavoidcarry-over,asalsoindicatedbyMirabellietal.[26]. Different spraying solvents were tested, namely a mixture of
acetonitrile and water(90:10 v/v),acetonitrile, methanol anda mixtureofmethanolwithwater(90:10v/v).Allsolventscontained 0.4% formic acid, 0.02mM Purine and 0.025mMHP-0921. Best result were achieved witha mixture of acetonitrileand water (90:10v/v)whensprayedataconstantvolumetricflowrateof3
m
L/minute, delivered by a syringe pump (Fusion 100, Chemyx,
Stafford, USA).The64lubricatedand 2blankfingerprintswere analyzedusingthesamesettings,butinsteadofdwelling,5scans of3mmweremeasuredwithineachfingerprint,usinga150
m
m/ secondscanrate(totalingto20s)andastepsizeof1mm.MSfull scandatawereacquiredwithAgilentMassHunterData Acquisi-tionsoftware(versionB.08.00).Beforedataanalysis,thefirstline of each analyzed fingerprint was removedfrom the results, as theseoftencontainedspectrawithlowintensities.Thiswaslikelya resultofsamplewetting,asdescribedbyBodzon-Kulakowskaetal. [29].Lubricantcomponentswereputativelyannotatedusingtheonline METLIN mass spectral metabolite database [30] and
comparisonwithpreviouslyobtainedresultsfromliterature. 2.5.Fingerprintimaging
Chemicalimages(1220mm)ofacyanoacrylatedblankand EXSlubricatedfingerprintwereacquiredusingthesame
param-eters as described above. The MS-inlet was cleaned after
acquisitionofeachchemicalimage.Imageswereacquiredusing a150
m
m/secondscanrate,resultingina30m
mpixelwidthand stepsize(totalingto400rows).DatawereconvertedtoimzML, using FireFly (v.3.0.1.1, Prosolia, Indianapolis, USA),and subse-quentlyanalyzedusingMSIReader(v1.01)[31].2.6.Statistics
DatawereconvertedtomzXMLandthe5000mostabundant
peakswerefilteredusingmsConvert[32].Dataweresubsequently
Table1
The32condomsusedinthisstudy,theabbreviationused,themanufacturerandtheircountryoforigin.
Condoms Abbreviation Manufacturer Country
BillyBoyExtraLubricated BBEL MapaHealthCare Germany
BalanceCondoom BC Condoom-anoniem Netherlands
BeppySoftComfort BSC Beppy Netherlands
DurexClassicalNatural DCN Durex UK
DurexExtraSafe DES Durex UK
DurexFeelingSensitive DFS Durex UK
DurexOrgasmic DO Durex UK
DurexPerforma DP Durex UK
DurexRealFeeling DRF Durex UK
DurexXLPower DXLP Durex UK
Euroglider EU AshaInternational/Euroglider Netherlands
EXSRegular EXS LTCHealthCare UK
FairSquaredOriginal FSO FairSquaredGMBH Germany GlydeUltraNaturelle GUN GlydeHealth Australia
JustSafeStandaard JSS Safe Netherlands
KruidvatClassic KC Kruidvat Netherlands
KruidvatExtra KE Kruidvat Netherlands
KruidvatSensationBanana KSb Kruidvat Netherlands KruidvatSensationChocolate KSc Kruidvat Netherlands KruidvatSensationStrawberry KSs Kruidvat Netherlands
KruidvatUltra KU Kruidvat Netherlands
LELOHEXCondooms LH Lelo Sweden
LevelPopular LP YourLevelsBV Netherlands
MoreAmoreSoftSkin MASS BizzyDiamondBV Netherlands
MySize MS R&SGermany Germany
MatesSKYNOriginal MSO LifestyleHealthcare Australia
ONNaturalFeeling ON R&SGermany Germany
PlayboyLubricated PL Playboy USA
PasanteNaturelle PN PasanteHealthcareLtd/Karex UK
Startex ST ForeSeeline Belgium
UniqPull UP UniqInternational Colombia
Wingman WI Wingman Netherlands
processedwithR(version3.4.2)usingRstudio(Version1.1.456), usingtheMALDIquantpackage[33].Massspectrawere square-roottransformedandnormalizedusingthetotalioncurrent(TIC). Afteraligning and averaging the spectra, peaks were detected usingthecorrespondingMALDIquantfunctions.Principal
compo-nent analysis (PCA) was then executed to reduce the data
dimensionality. After splitting the data in a 75% training and 25%testset,thefirst12PCs (explaining90% ofthecumulative proportionofvariance)wereusedtogeneratealineardiscriminant analysis(LDA)model,usingtheMASSpackage[34].Classification accuracywasevaluatedbygeneratingconfusionmatricesusingthe caretpackage[35].
3.Results
3.1.Detectionofcondomlubricants
As a firstscreening of thechemical components of condom lubricants,adetectionmethodwasdevelopedbasedontheanalyses of the 24-wells cyanoacrylated sample slides. Typical scans of lubricantsfrom4differentcondomsareshowninFigs.1–4.Ascanbe deducedfromthesespectra,distinctivepatterns,originatingfrom polymersthatmakeupalargepartofthelubricants,werefound. Closeranalysisofthedetectedm/zvaluesthatformtheseionseries leadtotheputativeannotationofthemajorcomponentsofthe condomlubricants(Table2).Alargecomponentofmanylubricants
appearedtobepoly(ethyleneglycol)(PEG),whereas m/zvalues correspondingtopolydimethylsiloxane(PDMS)weredetectedas well. In some lubricants, the polyethoxylated phenol nonionic surfactants octoxynol-9ornonoxynol-9, servingasspermicides, were observed (Table 2). Next to these chemical components, multipleionseriescorrespondingtothefattyalcoholethoxylates PEGdecyletherandPEGdodecyletherwerefound,oftenusedas non-ionicsurfactants(Table2)[36].Poly(propyleneglycol)(PPG) was putatively annotatedin someof the lubricants as well.In addition to the ion series resulting from the polymers largely presentincondomlubricants,afewmolecularionspecieswerealso detectedandputativelyannotated.Anexampleisthedetectionof benzocaine,alocalanestheticusedintwooftheDurexcondoms (Performa and Orgasmic) (Table 2). Furthermore, masses corre-spondingtoundecylamineanddodecylaminewereobserved.Asall samplesweresubjectedtocyanoacrylatefuming,acommonlyused detection technique for latent fingerprints within the forensic setting,amasscorrespondingtoacyanoacrylate(CA)fragmentwas alsofound(Table2).Finally,incontrols,exceptCA,noneoftheabove mentionedchemicalcomponentswerefound(datanotshown). 3.2.Differentiationofcondomlubricants
Todifferentiatebetweenthe32condoms,principalcomponent
analysis (PCA) and linear discriminant analysis (LDA) were
performed,as both areshown tobeeffectivein discriminating
Fig.1.SpectrumobtainedfromDESI-MSanalysisofspotsoflubricantfromBSCcondom,showingionseriescorrespondingtoPDMS()andnonoxynol-9[M+Na]+(*)and
[M+K]+
().
Fig.2.SpectrumobtainedfromDESI-MSanalysisofspotsoflubricantfromMSOcondom,showingionseriescorrespondingtopoly(ethyleneglycol)decylether(*), octoxynol-9[M+K]+()
and[M+Na]+
condom lubricants based on mass spectra [26,18,19]. PCA was performedtoreducedatadimensionality,usingthemassspectra acquiredfromthe32differentcondomlubricants(Fig.5).Using only the first two principal components (PCs), a distinction betweenmajorlubricantclassescouldalreadybemade.Togain moreinsight inwhichcomponentscan beusedtodifferentiate condomlubricants,theloadingsofthefirst5PCswereanalyzed (Fig.6).ThefirstPCcontainsm/zvaluescorrespondingtothe[M +Na]+and[M+K]+ionseriesofPEG,whereasinthesecondPCm/z
valuesofnonoxynol-9(both[M+Na]+and[M+K]+)are
incorpo-rated.InthethirdPC,putatively,octoxynol-9(both[M+Na]+and
[M+K]+)wasfoundtobethemajorcomponent.Undecylamineand
anunidentifiedm/zof848.6672arethemajorcontributorstothe
fourth PC. In the fifth PC, dodecylamine, PEG, PPG and an
unidentified m/zof 125.9863werethestrongest differentiating factors.
Next,lineardiscriminantanalysis(LDA)wasusedtogeneratea classificationmodelusingthefirst12PCs(explaining90%ofthe cumulativeproportionofvariance(Fig.S1)),basedonthetraining data(75%oflubricantdataobtainedfromanalysisofthe24-wells
slides). The generated model was subsequently evaluated by
classificationofthetestdata(25%oflubricantdataobtainedfrom analysisofthe24-wellsslides).Analysisoftheresultingconfusion matrixshowsthatthemodelisabletoclassifycondomlubricants withhighaccuracy(99.0%)(TableS1).Only1samplewaspredicted
incorrectly;asamplecontainingKSbwas predictedbythe PCA-LDAmodelasKSs(TableS1),whichbothoriginatefromthesame brandofflavoredcondoms(KruidvatSensations).Theselubricants likely contain the same basis, while different colorants and
flavorings are added. The fact that many components of the
lubricantarelikelytobeidentical,couldexplainthe missclassi-ficationofthegeneratedmodel.Subsequently,wegeneratedthe
PCA-LDA classification model based on the data from the
lubricatedcyanoacrylatedfingerprintsinthesamemanner.Using thisdata,anoverallaccuracyof90.9%wasachieved(TableS2).A fewmissclassificationwerepresentbutseemtobe comprehensi-ble,suchasthepredictionofDESasDCN(bothDurexcondoms)and the prediction of DP as DO (both Durex condoms that contain benzocaine).However,themodelperformedpoorforonespecific condomlubricant,namelyPL(sensitivityof40%).Furtheranalysis ofthePLdatarevealedthelowintensityofmanyofthechemical components, possibly explaining the poor performance of the modelinthiscase.
3.3.Imagingofcondomlubricantsinfingerprints
Visualisationofthepresenceofcondomtraceswithin finger-printswouldgreatlyenhancethestrengthof theevidence,asit bothestablishesthepresenceofthesuspectatthecrimesceneand contact with a condom. Therefore, full chemical images were
Fig.3.SpectrumobtainedfromDESI-MSanalysisofspotsoflubricantfromKCcondom,showingionseriescorrespondingtopoly(ethyleneglycol)dodecylether[M+Na]+
(*) andpoly(ethyleneglycol)decylether[M+Na]+().
Fig.4.SpectrumobtainedfromDESI-MSanalysisofspotsoflubricantfromDOcondom,showingionseriescorrespondingtopoly(ethyleneglycol)[M+H]+
(*),[M+K]+()
and [M+Na]+
().
obtainedfromfingerprintsthat handledan EXS condomand a blank(natural)fingerprint,bothtreatedwithCAfuming(Fig.7).As expected,CA(m/z556.1794)waspresentinbothfingerprints,and revealsthefrictionridgepatternofthefingerprintinbothcases
(Fig.7CandD).WhenrenderingthechemicaldistributionofPDMS (m/z 445.1200) for both fingerprints, only in the lubricated fingerprintadistinctiveimagewasacquired,thatwas absentin theblank(Fig.7AandB).A similarresultwasobtainedforthe
Table2
Detectedm/zvalues,theirputativeannotationandcorrespondingformula.
Putativeannotation Experimentalm/zvalues Formula n Ref
Benzocaine 166.0862 C9H11NO2 [M+H]+ – [30,37,18] Undecylamine 172.2058 C11H25N [M+H]+ – [30] Dodecylamine 186.2217 C12H27N [M+H]+ – [30] Ethylcyanoacrylate 556.1794 (C6H7NO2)n [M+H-C4H8N]+ n=5 – Poly(ethyleneglycol) 195.1226,239.1489,283.1753,327.1017,371.2279,415.2540, 459.2808,503.3059 H(C2H4O)nOH [M+H]+ n=4...11 [38,12] 217.1046,261.1308,305.1572,349.1836,393.2098,437.2359, 481.2622,525.2883,569.3160,613.3404,701.4077 H(C2H4O)nOH [M+Na]+ n=4...14 233.0785,277.1041,321.1298,365.1567,409.1832,453.2097, 497.2360,541.2620,585.2889,629.3150 H(C2H4O)nOH [M+K]+ n=4...13 Poly(ethyleneglycol) decylether 313.2348,357.2611,401.2873,445.3135,489.3393,533.3649, 577.3712 C10H21(C2H4O)nOH [M+Na]+ n=3...9 – Poly(ethyleneglycol) dodecylether 341.2662,385.2923,429.3183,473.3441,517.3681,561.3913, 605.4167 C12H25(C2H4O)nOH [M+Na]+ n=3...9 [39] Poly(propyleneglycol) 273.1674,331.2093,389.2514 H(C3H6O)nOH [M+Na]+ n=4...6 [38] 347.1857 H(C3H6O)nOH [M+K]+ n=5 Poly(dimethylsiloxane) 371.1013,445.1200,519.1382 (C2H6SiO)n [M+H]+ n=5...7 [38] 429.0882 (C2H6SiO)n [M+H-CH4]+ n=6 Octoxynol-9 449.2877,493.3136,537.3400,581.3661,625.3919,669.4184, 713.4439,757.4698 C14H21(C2H4O)nOH [M+Na]+ n=5....12 [38,40] 509.2876,553.3138,597.3399,641.3662,685.3925,729.4182, 773.4447,817.4707 C14H21(C2H4O)nOH [M+K]+ n=6...13 Nonoxynol-9 419.2772,463.3031,507.3291,551.3552,595.3811,639.4076 C15H23(C2H4O)nOH [M+Na]+ n=4...9 [38,40,12,21] 347.1982,391.2244,435.2509,479.2771,523.3033,567.3295, 611.3555,655.3815,699.4078,743.4352,787.4599 C15H23(C2H4O)nOH [M+K]+ n=2...12
Fig.5. Principalcomponentanalysisscoreplotbasedontheanalysisof32condomsandblank(n=12)usingthefirsttwoprincipalcomponents.Separateclusteringofseveral lubricantgroupsisobserved.
majorlubricantcomponentPEG(m/z585.2889,Fig.7EandF).PEG was foundto behighly abundantin the lubricatedfingerprint, whileonlyminorabundancewasfoundintheblank.Importantly, incaseofthelubricatedfingerprint,classificationusingasubsetof thedata(to getthesamenumber ofaveraged scans),correctly predictsthesourceofthelubricantasEXS(TableS3).
4.Discussion
Tothebestofourknowledge,thisisthefirststudydescribinga method for the differentiation of condom lubricants from CA
treated fingerprints using DESI-MS combined with a PCA-LDA
classificationmodel.Thegeneratedmodelshowedhighaccuracy forbothdirectanalysisofcondomlubricantspots(99.0%),aswell aslubricatedfingerprints(90.9%).Moreover,sincealargerangeof differentbrands and types of condoms, commonly soldin the
Netherlands, were analyzed, a representative database was
collected.
The detection, discrimination and visualization of condom derived traces from fingerprints is of significant evidential
importance in sexual assault cases, as it provides crucial
information onthe presence of a criminal at a crimescene as wellascontactwithacondomandtypeofcondomused,thereby greatlyincreasingthestrengthoftheevidence.Inpreviousstudies, ithasalreadybeenshownthatcondomlubricantscanbedetected anddiscriminated,solelyorwithinfingerprints,usingseveralMS techniques, includingDART-MS [18–20], MALDI-MS [21,22] and DESI-MS[25,26].However,inthesestudiestheeffectof cyanoac-rylate (CA)was not examined, while in forensics CAfuming is frequentlyperformedtovisualizefingerprintsastheyoftenappear
latent. We now show that the current described method is
compatiblewithcyanoacrylatefuming,renderingitmoresuitable for applicationto forensic casework.Additionally, full chemical imagescouldbeacquiredfromCAtreatedlubricatedfingerprints, showingthespatialdistributionoflubricantcomponentssuchas PEGandPDMSthroughoutthefingerprint,whichcanbecombined withclassificationofcondomlubricants.Thespatialinformation provided bychemical imaging, confirms thatthe lubricantwas
transferred by fingerprint contact as it links the presence of condomlubricanttothefingerprintridgedetail,makingitofmore evidentialvaluethanthesoleanalysisandcomparisonofcondom components.
MALDI-MS was previously shown to have the potential to
discriminatebetweendifferentcondombrandsortypes,combined withchemicalimaginginamultidisciplinaryanalyticalapproach, byBradshawetal.[22].However,theadvantageofusingDESI-MS ascomparedtoMALDI-MStechniques,isthatnomatrixorsample preparationisneededandanalysiscanbeperformedatambient pressure.DART-MSanalysisoffersstraightforwardanalysis with-outtheneedforsamplepreparation,andwasshowntobeableto achievehighclassificationaccuraciesbasedoncondomlubricant spectra [18–20], but lacks the capability to generate chemical images.WenowfoundthatDESI-MScombinestheeasyanddirect analysisofcondomlubricantsampleswiththeabilitytoperform
chemical imaging resulting in high accuracy detection and
discriminationofcondomtraces.AlthoughMALDI-MSiscapable of achieving higher spatial resolutions, the chemical images generatedusingDESI-MSshowclearridgedetail,whichwefound tobesufficientforthepurposeofthismethod.
Using the developed DESI-MS method, we found multiple
condom lubricant components. Among the most commonly
encountered compoundswereion seriescorrespondingtoPEG, PDMS,nonoxynol-9,octoxynol-9andPEGdodecylether.Basedon theloadingsofthefirstPCs,PEG,nonoxynol-9andoctoxynol-9 seemtobethemostdiscriminatorylubricantcomponents.Being
an essential part of many lubricant bases, PDMS, PEG and
nonoxynol-9 havebeenanalyzed fromcondomlubricanttraces using various analytical techniques, and have, not surprisingly, been included in many recent condom lubricant classification studies [18–20,25,26]. The detection of octoxynol-9in condom lubricantsislesscommonlyencountered,buthasbeendescribed by Thomas et al. [40] and Bradshaw et al. [21]. The putative annotationoftwofattyalcoholethoxylates(PEGdecyletherand PEGdodecylether),thatpossibleserveasethoxylate lubricants, are inagreementwithfindingsbyMirabelliet al.,whoalready mentionedthepossiblepresenceofethoxylatelubricantincertain
Fig.6.PlotoftheabsolutevaluesofcomponentloadingsoftheDESI-MSspectrafromcondomlubricantforthefirst5principalcomponents.
types of condoms [26]. The m/z values used for the putative annotationof poly(ethyleneglycol) dodecylether in ourstudy, correspondtopreviously describedpolymer fragmentsfroman unknownethoxylated polymer species by Mirabelli et al. [26]. Additionally,Musahetal.reportedthedetectionofoctylalcohol ethoxylatefromSkyncondoms,afterDART-MSanalysis[16].We
also detected m/z values corresponding to undecylamine and
dodecylamine, which, to our knowledge, are not commonly
detectedincondomlubricants,althoughoctylaminewasidentified in many of the previous studies, mainly used as emulsifier, dispersantorlubricant[16,25,18–20].
There are numerous alternative approaches available to
generateclassification models based onanalytical data. In our approach,weusedPCAasafirststep,toreducedata dimensional-ity,making the data easier to perceive. LDA was subsequently chosenasclassificationmethodasitshowedtobeaneasyandfast classificationmethod,whichhadalreadyproventobeeffectivein
discriminating condom lubricants based on mass spectra in
previousstudiesbyMaricetal.,Baumgartenetal.,andMirabelli
et al. [18,19,26]. In terms of classification accuracy based on lubricantspectra using DART-MS,Maricet al.achieved a98.7% accuracy based on classification of 90 lubricants to one of 12 distinctive groups [18], Baumgarten et al. acquired a 88.9% accuracywhenclassifying18differentlubricants[19],whileCoon etal. discriminated110condomtypesfrom16differentbrands witha 97.4% accuracy [20]. Classification of lubricants from10 differentcondomsusingDESI-MSbyMirabellietal.resultedina 94% accuracy [26]. Our results are largely in line with these previousstudies,aswegaineda99.0%accuracywhenanalyzing condomlubricantspots,anda90.9%accuracybasedonanalysisof lubricantcontainingfingerprints.Additionally,theseresultsshow thatthepresenceofCAdoesnotinterferewiththedetectionand discriminationofcondomlubricants,andhighaccuracyclassi fica-tion of CAfumed lubricanttraces using DESI-MSand PCA-LDA analysisisattainable.
Someofthemisclassificationsinourstudyseemtobecaused duetolubricantsoriginatingfromthesamecondombrand.When analyzinglubricantspots,asamplecontainingKSbwaspredicted
Fig.7.Chemicalimages(1220mm)showingthedistrubutionofPDMS(m/z445.1200,AandB),cyanoacrylate(m/z556.1794,CandD)andPEG(m/z585.2889,EandF) throughoutafingerprintcontainingEXSlubricant(A,CandE)andanaturalblank(B,DandF)fingerprint.
asKSs(bothKruidvatcondoms),whileinfingerprintscontaining lubricants,DESwaspredictedasDCN(bothDurexcondoms)and
DP was predicted as DO (both Durex condoms that contain
benzocaine).Themisclassificationofcondomlubricantoriginating fromtwo differentDurex sources was also experienced in one occasion by Mirabelli et al. [26], likely being the result of similaritiesbetweencondomlubricantsoriginatingfromthesame
brand. This was shown by Maric et al. and Coon et al., who
classifiedcondomlubricantstoamajorlubricantgroup/brandwith highaccuracy[18,20].Predictingthecondomlubricanttracesby brandonly,insteadofbrandandtype,wouldpresumablyleadtoan increased classification accuracy in our study as well. For one particular condom (PL), we found a low sensitivity (40%) in lubricatedfingerprints,whichseemedtobetheresultoflowion intensities, possibly explaining the poor performance of the statisticalmodelinthiscase.
As the majorcomponents of condom lubricants are known
contaminantsinmassspectrometry[38],weencountered carry-overproblemsduringmethoddevelopmentandoptimization,that weresimilartotheeffectsdescribedbyMirabellietal.[26].Intheir study,itwasfoundthatthemostrelevantparametersdetermining the‘memoryeffect’werethedistancebetweenthespraytipand iontransferlineandbetweentheiontransferlineandsample.Too shortdistancesresultedincontaminationoftheiontransferline, assamplematerialcouldbesuckedintotheMSinlet[26].Indeed, we also found that increasing the ion transfer line-to-surface distanceandspraytip-to-iontransferlinedistance,togetherwith cleaningtheMSinletafter3–4samples,resultedinavoidanceof samplecarry-over,indicatingthatthesearecrucialsettingsand actionsforreliableresultswhenanalyzingcondomlubricanttraces withDESI-MS.Also,whenimaging lubricatedfingerprintsusing DESI-MS,wefoundthathighamountsofcondomlubricantinthe fingerprintsdidnotgeneratehighqualitychemicalimages,dueto
a decrease in clear ridge detail as a consequence of high
abundances of PDMS and PEG ion signals. However, the
classification model still predicted the source of the lubricant correctly,indicatingthatadiscriminationcouldstillbemade.
In this paper, we solely focused on theanalysis of condom lubricant traces in CA treated fingerprints. However, the main componentsofthesecondomlubricants,suchasPEGandPDMS, canalsobefoundinmanypersonalcareproducts[15].Asaresult, analysisoffingerprintsthatpossiblycontaintracesofanyofthese personalcareproducts,mayleadtomisclassifications.Although theability todiscriminatebetweenpersonal careproductsand condomlubricantsinfingerprintswasnotanalyzedinthepresent study,arecentstudyperformedbyMoustafaandBridgeshowed thatdiscriminationbetweentheseclassesofproductsispossible using DART-MS and LDA [15]. The addition of discriminating factorsfromotherclassesofpersonalcareproductstothecurrent developedmodelwouldfurtherincreasetheforensicapplicability
of the generated method. Furthermore, we only measured
fingerprintswithcondomlubricanttracesfromglasssubstrates, while in practice, fingerprints can be found on all available substrates. Furtheroptimization of the analysis of fingerprints containingcondomlubricanttracesonseveraldifferentsubstrates wouldalsobenefitthedevelopedmethod.Indeed,Mirabellietal.
showed that chemical analysis and imaging of fingerprints
containing condom lubricant is possible on metal and paper
surfaces[25,26].However,spectraobtainedfrompapersurfaces hadlowersignalintensitiesduetosorptioneffects,andawash-out effectwasencounteredwhenanalyzingonmetalsurfaces[25,26]. 5.Conclusion
We developed a DESI-MS method for the detection and
discriminationofcondomlubricanttracesfromfingerprintsthat,
combined with a PCA-LDA classification model, has an overall accuracyof 90.9%andis compatiblewithCAfuming,making it moreapplicableforforensiccasework.Additionally,fullchemical imagesoffingerprintscontainingcondomlubricanttracescouldbe acquired,visualizingthespatialdistributionofcondomlubricant compounds,suchasPDMSandPEG.Thisconfirmsthatthecondom lubricantisoriginatingfromthefingerprintandnotthesubstrate, therebyincreasingevidentialstrength.Theseresultsarepromising leadsforfurtherdevelopmentofDESI-MSmethodstoqualitatively
analyze exogenous compounds from fingerprints for use in
forensicscience.
CRediTauthorshipcontributionstatement
WardvanHelmond:Conceptualization,Methodology,Formal analysis,Investigation,Writing-originaldraft,Visualization.Mark P.V.Begieneman:Conceptualization,Methodology,Investigation, Writing - original draft. Roos Kniest: Investigation, Writing -review&editing.MarceldePuit:Conceptualization,Methodology, Writing - review & editing, Project administration, Funding acquisition.
DeclarationofCompetingInterest Therearenoconflictstodeclare. Acknowledgement
WvHacknowledgesaRAAK-PROresearchgrant(no. 2014-01-124PRO),theNetherlands.
AppendixA.Supplementarydata
Supplementarymaterialrelatedtothisarticlecanbefound,in theonline version,at doi:https://doi.org/10.1016/j.forsciint.2019. 110005.
References
[1]WorldHealthOrganization,GlobalandRegionalEstimatesofViolenceagainst Women:PrevalenceandHealthEffectsofIntimatePartnerViolenceand Non-PartnerSexualViolence,WorldHealthOrganization,2013.
[2]R.D.Blackledge,ForensicAnalysisontheCuttingEdge:NewMethodsforTrace EvidenceAnalysis,JohnWiley&Sons,2007.
[3]K.C.Davis,T.J.Schraufnagel,W.H.George,J.Norris,Theuseofalcoholand condomsduringsexualassault,Am.J.MensHealth2(3)(2008)281–290. [4]R.D. Blackledge, Condom traceevidence: a newfactor insexual assault
investigations,FBILEnforcementBull.65(12)(1996).
[5]A.Raj,E.Reed,E.Miller,M.R.Decker,E.F.Rothman,J.G.Silverman,Contextsof condomuseandnon-condomuseamongyoungadolescentmaleperpetrators ofdatingviolence,AIDSCare19(8)(2007)970–973.
[6]E.N.O’Neal,S.H.Decker,C.Spohn,K.Tellis,Condomuseduringsexualassault,J. ForensicLeg.Med.20(6)(2013)605–609.
[7]P.Brauner,N.Gallili,Acondom—thecriticallinkinarape,J.ForensicSci.38(5) (1993)1233–1236.
[8]R.Blackledge,Collectionandidentificationguidelinesfortracesfromlatex condomsinsexualassaultcases,CrimeLab.Digest21(4)(1994)57–61. [9]R.Blackledge,M.Vincenti,Identificationofpolydimethylsiloxanelubricant
tracesfromlatexcondomsincasesofsexualassault,J.ForensicSci.Soc.34(4) (1994)245–256.
[10]G.P. Campbell, A.L. Gordon, Analysis of condom lubricants for forensic casework,J.ForensicSci.52(3)(2007)630–642.
[11]T.Coyle,N.Anwar,Anovelapproachtocondomlubricantanalysis:in-situ analysisofswabsbyFT-RamanspectroscopyanditseffectsonDNAanalysis, Sci.Justice49(1)(2009)32–40.
[12]P.Maynard,K.Allwell,C.Roux,M.Dawson,D.Royds,Aprotocolfortheforensic analysisofcondomandpersonallubricantsfoundinsexualassaultcases, ForensicSci.Int.124(2-3)(2001)140–156.
[13]R.A.Musah,A.L.Vuong,C.Henck,J.R.Shepard,Detectionofthespermicide nonoxynol-9viaGC-MS,J. Am.Soc.Mass Spectrom.23(5) (2012)996– 999.
[14]F.Burger,M.Dawson,C.Roux,P.Maynard,P.Doble,P.Kirkbride,Forensic analysis of condom andpersonal lubricants bycapillary electrophoresis, Talanta67(2)(2005)368–376.
[15]Y. Moustafa,C.M. Bridge,Distinguishing sexuallubricants from personal hygieneproductsforsexualassaultcases,ForensicChem.5(2017)58–71. [16]R.A.Musah,R.B.Cody,A.J.Dane,A.L.Vuong,J.R.Shepard,Directanalysisinreal
time mass spectrometry for analysis of sexual assault evidence, Rapid Commun.MassSpectrom.26(9)(2012)1039–1046.
[17]G.Proni,P.Cohen,L.-A.Huggins,N.Nesnas,Comparativeanalysisofcondom lubricantsonpre&post-coitalvaginalswabsusingAccuTOF-DART,Forensic Sci.Int.280(2017)87–94.
[18]M.Maric,L.Harvey,M.Tomcsak,A.Solano,C.Bridge,Chemicaldiscriminationof lubricantmarketingtypesusingdirectanalysisinrealtimetime-of-flightmass spectrometry,RapidCommun.MassSpectrom.31(12)(2017)1014–1022. [19]B.Baumgarten,M.Maric,L.Harvey,C.M.Bridge,Preliminaryclassification
schemeofsiliconebasedlubricantsusingDART-TOFMS,ForensicChem.8 (2018)28–39.
[20]A.M.Coon,S.Beyramysoltan,R.A.Musah,Achemometricstrategyforforensic analysisofcondomresidues:identificationandmarkerprofilingofcondom brandsfromdirectanalysisinrealtime-highresolutionmassspectrometric chemicalsignatures,Talanta194(2019)563–575.
[21]R.Bradshaw,R.Wolstenholme,R.D.Blackledge,M.R.Clench,L.S.Ferguson,S. Francese,Anovelmatrix-assistedlaserdesorption/ionisationmass spectrom-etry imaging based methodology forthe identification of sexualassault suspects,RapidCommun.MassSpectrom.25(3)(2011)415–422.
[22]R.Bradshaw,R.Wolstenholme,L.S.Ferguson,C.Sammon,K.Mader,E.Claude, R.D. Blackledge, M.R. Clench, S. Francese, Spectroscopic imaging based approachforcondomidentificationincondomcontaminatedfingermarks, Analyst138(9)(2013)2546–2557.
[23]A.Bodzon-Kulakowska,P.Suder,Imagingmassspectrometry: instrumenta-tion,applications,andcombinationwithothervisualizationtechniques,Mass Spectrom.Rev.35(1)(2016)147–169.
[24]F.Green,T.Salter,I.Gilmore,P.Stokes,G.O’connor,Theeffectofelectrospray solventcompositionondesorptionelectrosprayionisation(DESI)efficiency andspatialresolution,Analyst135(4)(2010)731–737.
[25]M.F.Mirabelli,A.Chramow,E.C.Cabral,D.R.Ifa,Analysisofsexualassault evidencebydesorptionelectrosprayionizationmassspectrometry,J.Mass Spectrom.48(7)(2013)774–778.
[26]M.F. Mirabelli,D.R.Ifa,G.Sindona, A.Tagarelli,Analysisofsexualassault evidence:statisticalclassificationofcondomsbyambientmassspectrometry, J.MassSpectrom.50(5)(2015)749–755.
[27]C.Champod,C.J.Lennard,P.Margot,M.Stoilovic,FingerprintsandOtherRidge SkinImpressions,CRCPress,2016.
[28]S.M.Bleay,R.S.Croxton,M.DePuit,FingerprintDevelopmentTechniques, WileyOnlineLibrary,2018.
[29]A.Bodzon-Kulakowska,A.Drabik,J.Ner,J.H.Kotlinska,P.Suder,Desorption electrosprayionisation(DESI)forbeginners–howtoadjustsettingsfortissue imaging,RapidCommun.MassSpectrom.28(1)(2014)1–9.
[30]C.A.Smith,G.O’Maille, E.J.Want,C. Qin,S.A.Trauger,T.R.Brandon,D.E. Custodio, R. Abagyan, G. Siuzdak, METLIN: a metabolite mass spectral database,Ther.DrugMonit.27(6)(2005)747–751.
[31]M.T. Bokhart, M. Nazari, K.P. Garrard, D.C. Muddiman, MSiReader v1.0: evolvingopen-sourcemassspectrometryimagingsoftwarefortargetedand untargetedanalyses,J.Am.Soc.MassSpectrom.29(1)(2018)8–16. [32]M.C.Chambers,B.Maclean,R.Burke,D.Amodei,D.L.Ruderman,S.Neumann,
L.Gatto,B.Fischer,B.Pratt,J.Egertson,Across-platformtoolkitformass spectrometryandproteomics,Nat.Biotechnol.30(10)(2012)918. [33]S.Gibb,K.Strimmer,MALDIquant:aversatileRpackagefortheanalysisof
massspectrometrydata,Bioinformatics28(17)(2012)2270–2271. [34]B.D.Ripley,W.N.Venables,ModernAppliedStatisticswithS,vol.537,Springer,
NewYork,2002.
[35]M.Kuhn,BuildingPredictiveModelsinRUsingtheCaretPackage28(5):26, (2008),doi:http://dx.doi.org/10.18637/jss.v028.i05.
[36]S.S.Talmage,EnvironmentalandHumanSafetyofMajorSurfactants:Alcohol EthoxylatesandAlkylphenolEthoxylates,CRCPress,1994.
[37]S.E.Spencer,S.Y.Kim,S.B.Kim,K.A.Schug,Matrix-assistedlaserdesorption/ ionization–timeofflight-massspectrometryprofilingoftraceconstituentsof condomlubricantsinthepresenceofbiologicalfluids,ForensicSci.Int.207(1– 3)(2011)19–26.
[38]B.O.Keller,J.Sui,A.B.Young,R.M.Whittal,Interferencesandcontaminants encounteredinmodernmassspectrometry,Anal.Chim.Acta627(1)(2008) 71–81.
[39]R.M.Facino,M.Carini,P.Minghetti,G.Moneti,E.Arlandini,S.Melis,Direct analysisofdifferentclassesofsurfactantsinrawmaterialsandinfinished detergent formulations by fast atom bombardment mass spectrometry, Biomed.Environ.MassSpectrom.18(9)(1989)673–689.
[40]J.J.Thomas,Z.Shen,R.Blackledge,G.Siuzdak,Desorption–ionizationonsilicon massspectrometry:anapplicationinforensics,Anal.Chim.Acta442(2) (2001)183–190.