Thesis (MSc)--Stellenbosch University, 2012 / ENGLISH ABSTRACT: The aroma and flavour of wine are important aspects that form the basis for consumers’ organoleptic experience of wine. Therefore, an understanding of the chemical composition of wine aroma is of major importance, to establish possible links between wine chemistry, sensory attributes and consumer preference for a product. For this purpose analytical chemistry and multivariate techniques are indispensable tools for the metabolic profiling of wine. Chenin blanc is one of the most important South African export white wine varieties. However, despite its importance, very limited profiling of Chenin blanc aroma compounds has been done and information is restricted to isolated and dated reports on a few chemical compounds only. Therefore, the overall aim of this study was to obtain an in-depth view of the volatile chemical profile of this cultivar. The first task was to perform targeted volatile metabolic profiling of the three dry and offdry Chenin blanc styles, fresh and fruity, rich and ripe unwooded and rich and ripe wooded. To this end, a new, simple and robust liquid-liquid extraction technique using dichloromethane was developed and validated for extraction of analytes prior to gas chromatography flame ionization detection (GC-FID) analysis, to quantify 57 analytes in one rapid analytical procedure. This method was applied to profile 48 Chenin blanc wines. Very successful discrimination between the three styles, using the quantified volatile compounds, was obtained with two multivariate methods. These were partial least squares regression-discriminant analysis, (PLS-DA), as well linear-DA, using best subset selection for identifying the most important variables. According to the classification models, a higher content of maturation derived, malolactic fermentation derived and wood derived compounds were predominantly characteristic of the wooded wines. Higher content of some terpenes and ethyl esters were predominantly associated with the rich and ripe unwooded style Chenin blanc wines, while the fresh and fruity style were generally characterized by high levels of acetate esters. Secondly, untargeted analysis of 21 wines was done with gas chromatography mass spectrometry (GC-MS). Mathematical chromatography, using PARAllel FACtor analysis (PARAFAC and PARAFAC2), was applied to the GC-MS data for resolution of the complex chromatographic results by multi-way modeling, and to derive unbiased multivariate classification models of the three styles. This approach provided excellent style differentiation, without the arduous task of analysis of numerous standards and setting up of calibration curves, required by the targeted approach described above. Additionally, the data generated during this study will form part of the current South African wine aroma database, which does not contain any data regarding Chenin blanc at present. / AFRIKAANSE OPSOMMING: Die aroma en geur van wyn is belangrike aspekte aangesien dit die basis vorm van die wynverbruiker se organoleptiese ervaring van die produk. Derhalwe, is ‘n deeglike kennis van die chemiese samestelling van wynaroma baie belangrik, ten einde die korrelasies tussen wynchemie, sensoriese eienskappe en verbruikersvoorkeure te bepaal. Vir hierdie doel, is die insameling van analitiese chemiese data, tesame met multi-veranderlike tegnieke om relevante inligting uit die data te onttrek, onmisbaar vir die metaboliese profilering van wyn. Chenin blanc is huidiglik een van Suid-Afrika se belangrikste uitvoer wit wynvariëteite. Ten spyte hiervan, is daar tot hede egter baie min profilering van Chenin blanc se aromakomponente gedoen en die beskikbare inligting is beperk tot geïsoleerde en verouderde navorsingsbevindinge. In die lig van bogenoemde, is die oorkoepelende motivering vir hierdie studie dus om die vlugtige chemiese komponente se profiel in Chenin blanc wyn, in diepte te bepaal. Die eerste taak was die geteikende bepaling van die metaboliese profiel van drie droë of halfdroë Chenin blanc wynstyle, nl. vars en vrugtig, ryk en ryp ongehout, asook ryk en ryp gehout. Om dit te bereik, is ‘n nuwe eenvoudige en robuuste vloeistof-vloeistof ekstraksieprosedure met dichlorometaan ontwikkel, wat analise met gaschromatografie – vlamionisasie deteksie (GC-FID) voorafgaan, om die konsentrasies van 57 komponente in een vinnige analise te bepaal. Hierdie metode is gebruik om 48 Chenin blanc wyne te profileer. Deur gebruik te maak van multi-veranderlike data analitiese tegnieke, is die gekwantifiseerde vlugtige komponente data gebruik in diskriminant analise. Baie suksesvolle onderskeid tussen die drie style, is verkry deur gebruik te maak van twee multi-veranderlike metodes, naamlik: parsiële kleinste kwadrate regressie, diskriminant analise, asook liniêre diskriminant analise. Vir laasgenoemde analise, is die seleksie van die mees belangrike veranderlikes met beste sub-groep regressie bepaal. Volgens hierdie klassifikasie modelle is ‘n hoër inhoud van veroudering-, appelmelksuurgisting- en houtverwante aroma komponente baie kenmerkend in die houtbehandelde wyne. Hoër vlakke van sommige terpene en etiel esters was kenmerkend met betrekking tot ryk en ryp ongehoute Chenin blanc style, terwyl die vars en vrugtige style meer gekenmerk was met hoë vlakke van asetaat esters. Tweedens is ‘n seleksie van 21 wyne geanaliseer deur gebruik te maak van gaschromatografie – massa spektrometrie (GC-MS) in ‘n ongeteikende metaboliese profileringsbenadering. Wiskundige chromatografiese metodes, spesifiek, parallelle faktor analise (PARAllel FACtor analysis (PARAFAC and PARAFAC2)), was voorts gebruik om komplekse chromatogramme te prosesseer met wiskundige, multi-vlak modellering. Met hierdie nie-selektiewe benadering, is ook suksesvolle klassifikasiemodelle gegenereer vir die diskriminasie tussen die drie verskillende Chenin blanc style. Die voordeel van die ongeteikende GC-MS analise, gekoppel met die data hanterings- en prosesseringsprotokols in hierdie studie gebruik, is dat die arbeidsintensiewe taak om kalibrasiekurwes op te stel vir elke individuele komponent, soos wat vereis word in die geteikende benadering, nie nodig is nie. Die data wat ingewin is gedurende hierdie studie, sal ook bygevoeg word tot ‘n bestaande Suid-Afrikaanse aroma databasis, wat tans geen data aangaande Chenin blanc wyn bevat nie.