Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
The Spiced project: Seasonable doubt...
Saskia van Ruth ([email protected])
Partners Spiced WP4: BfR, BIOR, Fuchs, DLO-RIKILT
Are we dining on deception?
3
Souls for sale?
Food product adulteration and counterfeiting is a thriving multi-billion euro global industry. It is highly profitable and the risks of significant legal consequences are low.
Fraud reports in three databases
0 50 100 150 200
Herbs and spices
Olive oil
Fish and fish products
Milk and milk products
Meat and edible offal
Vegetable fats and oils…
Nuts, nut products and…
Honey and royal jelly
Other seafood…
Fraud cases 2008 - 2013
Food Fraud Food Shield RASFF
Food Fraud
Food Defence
Food Safety
Food Quality
Ideologically driven Motivation is ‘HARM’
Food borne illness
Economically driven
Motivation is ‘GAIN’ Intentional Adulteration
Unintentional / Accidental Contamination
Petra Wissenburg, Danone Corporate Quality & Food Safety
EU project SPICED fraud detection methods
The aim?
• To develop rapid and cost-efficient methodologies for detection of contamination (adulteration) of spices/herbs with chemical agents
• Aiming at early detection of anomalies in view of food integrity, i.e. signalling/flagging for unusual samples (authentic vs adulterated)
By which means?
• By exploring spectroscopic and spectrometric fingerprinting techniques
9
Development of broad anomaly fingerprint tests
0
20
40
60
80
100
120
140
160
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
0
20
40
60
80
100
120
140
160
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
10
Overview of activities
11
Method set-up
for spices
Data genuine samples
Data self-adul-
terated samples
Data
bases spices
Broad Anomaly models/ compa-
rison
Herbs
• Basil
• Thyme
• Oregano
Development of techniques for broad anomaly testing of spices
Spices
• Pepper
• Chili paprika
• Nutmeg
12
Selected analytical techniques
Spectroscopy
• FTIR
• NMR
Mass spectrometry
• PTR-MS
• DIMS
• (LC)-HRMS
• ICP-MS
16
Approach: one class classification model
Normal variation, Database samples (50)
Samples with deviations from what is considered normal
Building the best statistical model for discrimination
Future samples to be tested
17
Approach
50 Authentic baseline samples of each spice were analysed with each method
One class classification models were built based on those authentic samples
Inferior/adulterant material, and mixtures of authentic and inferior/adulterant material were analysed
The models were challenged with the authentic and modified sample material
Optimised classification models of the various methods were compared to select most promising approaches
18
Nutmeg
1
10
100
1000
10000
100000
22
32
42
52
62
72
82
92
10
2
11
2
12
2
13
2
14
2
15
2
16
2
17
2
18
2
19
2
20
2
21
2
Inte
nsi
ty`[
pp
bv]
Mass [amu]
Nutmeg spectrum Terpenes: pinenes, carenes, limonene Myrcene, cymene, sabinene, camphene
1,2
-cin
eole
safr
ole
meth
yle
ugenol
asaro
ne
myristicin
acetic a
cid
1-b
uta
nol
One class classification result: Nutmeg
Baseline
sam
ple
s
Adultera
tions
Challenge
bla
nks
Cla
ss d
ista
nce
• SIMCA model, 7 factors • Normalised, 10log scaled
data
Classification results:
Prediction Authentic Adulterated % correct
Baseline samples 47 3 94% Challenge samples:
Authentic 5 0 100%
Adulterated 4 38 90%
Best models: discrimination 100% real nutmeg and own material mixed samples
Volatiles PTR-MS: • kNN model
• 100% real nutmeg and 100% correct prediction for own material mixed samples
• Estimated limit of detection 7%
Non-volatiles FI-ESI-MS: • PCA Q values
• Similar success rate predictions
• Limit of detection 30%
Successful classification PTR-MS - nutmeg
0
20
40
60
80
100
Authentic training Authenticchallenge
Adulterantchallenge
Mixed adulterantchallenge
[%]
24
Conclusions
Broad anomaly testing approaches have been developed for a variety of spices and herbs and various fraud issues in Spiced
For each spice/herb at least a few approaches were able to distinguish well between authentic and adulterated materials
Because of the one class classifier approach, the methods are expected to be able to detect a variety of other deviations from normal, authentic sample material
To increase robustness the training sets of authentic material could be expanded to cover more “special but natural” variation of the spice and herb materials
25
More presentations on the topic to follow
Bettina Horn (BfR): More fingerprinting methods, especially on paprika powders
Vadims Bartkevičs: The targeted approach
26