Chemometrics-assisted simultaneous voltammetric determination of ascorbic acid,uricacid,dopamine and nitrite: Application of non-bilinear voltammetric data for exploiting first-order advantage

GHOLIVAND M; JALALVAND A; GOICOECHEA H C; SKOV T

TALANTA

ELSEVIER SCIENCE BV

Lugar: Amsterdam; Año: 2014 vol. 119 p. 553 - 563

0039-9140

For the first time,severalmultivariatecalibration(MVC)modelsincludingpartialleastsquares-1(PLS-1), continuum powerregression(CPR),multiplelinearregression-successiveprojectionsalgorithm(MLR- SPA),robustcontinuumregression(RCR),partialrobustM-regression(PRM),polynomial-PLS(PLY-PLS), spline-PLS(SPL-PLS),radialbasisfunction-PLS(RBF-PLS),leastsquares-supportvectormachines (LS-SVM),wavelettransform-artificial neuralnetwork(WT-ANN),discretewavelettransform-ANN (DWT-ANN),andbackpropagation-ANN(BP-ANN)havebeenconstructedonthebasisofnon-bilinear first ordersquarewavevoltammetric(SWV)dataforthesimultaneousdeterminationofascorbicacid (AA), uricacid(UA),dopamine(DP)andnitrite(NT)ataglassycarbonelectrode(GCE)toidentifywhich techniqueoffersthebestpredictions.Thecompositionsofthecalibrationmixtureswereselected according toasimplexlatticedesign(SLD)andvalidatedwithanexternalsetofanalytes'mixtures.An asymmetric leastsquaressplinesregression(AsLSSR)algorithmwasappliedforcorrectingthebaselines. A correlationoptimizedwarping(COW)algorithmwasusedtodataalignmentandlackofbilinearitywas tackled bypotentialshiftcorrection.Theeffectsofseveralpre-processingtechniquessuchasgenetic algorithm (GA),orthogonalsignalcorrection(OSC),meancentering(MC),robustmediancentering (RMC), waveletdenoising(WD),andSavitsky?Golay smoothing(SGS)onthepredictiveabilityof the mentionedMVCmodelswereexamined.Thebestpreprocessingtechniquewasfoundforeach model. Accordingtotheresultsobtained,theRBF-PLSwasrecommendedtosimultaneouslyassaythe concentrations ofAA,UA,DPandNTinhumanserumsamples