HTC-15

HTC-15 - Abstract

Abstract Title: Can Gas Chromatography – Olfactometry Determine the Importance of Volatile Organic Chemical to Food and Beverage Odour?
Abstract Type: Seminar
Presenter Name: Mr Lewis Jones
Co-authors:Dr Richard Haydock
Mrs Kathleen Pinfold
Mr James Addison
Company/Organisation: Sensient Flavors
Session Choice: Food/Drink

Abstract Information :

Gas Chromatography - Olfactometry (GC-O) has been established for over 30 years. GC-O is used for the bioactivity guided identification of odor active volatile organic compound (odorants). However, there is little information of the techniques validity. Within this work the validity of GC-O has been investigated showing: (i) that humans are very imprecise GC detectors; (ii) that the methods used with the GC-O technique to prioritise odorants, such as Aroma Extract Dilution Analysis (AEDA), are very poor predictors of an odorants importance to a food or beverage. It is therefore recommended that when using GC-O multiple analyses should be carried out for odorant identification, and when using GC-O for prioritisation of odorants, other measures need to be taken into account.

To validate humans as GC detectors the probability of peak detection for two odorants, hexanal and 3-methylbutanal, over a range of concentration were studied. Using a probit model to transform the data, an estimate of the mean concentration for a 95% probability of odor detection was calculated with 95% confidence limits. For both odorants this "limit of detection" was unexpectedly high (95 and 28.6 mg / L respectively) with imprecision (upper 95% confidence limit 921 and 110 mg / L, lower 95% confidence limit 38.9 and 15.6 mg / L respectively).

To investigate the GC-O technique's ability to prioritise odorants, a meta-analysis of a selection of publications were carried out. For the meta-analysis, GC-O results for each odorant were used to predict an odorants odour-activity-value (OAV). The OAV is the ratio of the concentration of an odorant in the food and its odour detection threshold in a suitable matrix. OAV is assumed to relate to an odorants overall importance in a food. Using a simple linear model, results from the prediction were very poor with root mean square deviation (RSME) 1.14. However, by using additional data, such as an odorants vapour pressure and air/water partition coefficient, with a machine learning model, it was found that prediction improved with RMSE 0.58. Interestingly, within the model the best single predictor of OAV was linear retention index on a standard polar column. This shows that a combination of analytical measures is actually better at predicting an odorants importance to food than results from labour intensive GC-O methodologies.