Mycotoxin contamination of cereal grains can be detected and quantified using complex
extraction procedures and analytical techniques. Normally, the grain odour, i.e. the
presence of non-grain volatile metabolites, is used for quality classification of
grain. We have investigated the possibility of using fungal volatile metabolites as
indicators of mycotoxins in grain. Ten barley samples with normal odour, and 30 with
some kind of off-odour were selected from Swedish granaries. The samples were evaluated
with regard to moisture content, fungal contamination, ergosterol content, and levels
of ochratoxin A (OA) and deoxynivalenol (DON). Volatile compounds were also analysed
using both an electronic nose and gas chromatography combined with mass spectrometry
(GC-MS). Samples with normal odour had no detectable ochratoxin A and average DON
contents of 16 microg kg(-1) (range 0-80), while samples with off-odour had average
OA contents of 76 microg kg(-1) (range 0-934) and DON contents of 69 microg kg(-1)
(range 0-857). Data were evaluated by multivariate data analysis using projection
methods such as principal component analysis (PCA) and partial least squares (PLS).
The results show that it was possible to classify the OA level as below or above the
maximum limit of 5 microg kg(-1) cereal grain established by the Swedish National
Food Administration, and that the DON level could be estimated using PLS. Samples
with OA levels below 5 microg kg(-1) had higher concentration of aldehydes (nonanal,
2-hexenal) and alcohols (1-penten-3-ol, 1-octanol). Samples with OA levels above 5
microg kg(-1) had higher concentrations of ketones (2-hexanone, 3-octanone). The GC-MS
system predicted OA concentrations with a higher accuracy than the electronic nose,
since the GC-MS misclassified only 3 of 37 samples and the electronic nose 7 of 37
samples. No correlation was found between odour and OA level, as samples with pronounced
or strong off-odours had OA levels both below and above 5 microg kg(-1). We were able
to predict DON levels in the naturally contaminated barley samples using the volatile
compounds detected and quantified by either GC-MS or the electronic nose. Pentane,
methylpyrazine, 3-pentanone, 3-octene-2-ol and isooctylacetate showed a positive correlation
with DON, while ethylhexanol, pentadecane, toluene, 1-octanol, 1-nonanol, and 1-heptanol
showed a negative correlation with DON. The root mean square error of estimation values
for prediction of DON based on GC-MS and electronic nose data were 16 and 25 microg
kg(-1), respectively.