Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
SUPPLEMENTARY INFORMATION
2p or not 2p: Tuppence-based SERS for the detection of illicit materials
Figure S1. Deposition of silver (Grey target) demonstrated on a post-1992 2p coin.
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Figure S2. Raman spectrum generated when Mephedrone (1x10-4M) was spotted onto the bare copper coin.
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Figure S3. Representative SERS spectra of the drugs ketamine, cocaine and amphetamine acquired from the silver surface.
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
All vibrational assignments were carried out using: G. Socrates. Infrared and Raman Characteristic Group Frequencies: Tables and Charts,
Wiley: New York, U.S.A. 2001.
Table S1. Tentative SERS vibrational assignments for Mephedrone
Raman Shift (cm-1)
Vibration Assignment
418.6
530
602.2
698.5
741.5
799.5
899
991.3
1030
1103
1215
1292
1354
1459
1509
1558
1604
1657
CNC def
CCC ring or CO deformation
In plane ring deformation
Asym CNC stretch or ring vib para substituted benzene
Asym CNC stretch or ring vib para substituted benzene
C-H out-of-plane bending vibration
Unassigned
Unassigned
CH in plane deformation (benzene)
Sym CNC str
CN Stretch or para disubstituted ring vibration
Unassigned
CH3CO- (Symmetrical CH3 deformation)
Unassigned
NH deformation secondary Amine
Benzene derivative C=C str
Benzene derivative C=C str or C=C conjugated with C=O
C=O stretch
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Table S2. Tentative SERS vibrational assignments for MDMA
Raman Shift (cm-1)
315.2
425
466.5
517.3
567.8
617.8
729.2
814.6
860
967.6
1050
1108
1149
1201
1246
1337
1429
1465
1514
1558
Vibrational Assignment
Unassigned
Unassigned
Unassigned
CCC ring
CCC ring
CCC ring
Unassigned
OCCO
CNC
COC
Unassigned
CH2
CNC
Unassigned
CH
Unassigned
CH3/CH2 scissors
CH3/CH2 scissors
COC
Unassigned
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Table S3. Tentative SERS vibrational assignments for MDAI
Raman Shift (cm-1)
453.8
472.9
514.2
611.6
744.6
814.6
863
917
961.6
1033
1103
1212
1337
1370
1476
1530
1569
1700
Vibrational Assignment
Tetrasubstituted Benzene
Tetrasubstituted Benzene
Tetrasubstituted Benzene
Tetrasubstituted Benzene
Unassigned
sym C-O-C str
C-H out of plane deformation associated with benzene ring
Unassigned
cyclo pentene ring vibration
1,3-dioxolane ring vibration
primary amine vibration or 1,3 dioxolane ring vibration
Unassigned
Unassigned
C=C str (dioxolane)
1,2,4,5-tetrasubstituted benzene vib
1,2,4,5-tetrasubstituted benzene vib
Unassigned
Unassigned
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Table S4. Tentative SERS vibrational assignments for Ketamine
Raman Shift (cm-1)
360.6
434.6
526.8
611.6
692.3
741.5
790.3
841.9
922.9
1030
1076
1108
1178
1238
1286
1337
1440
1533
1590
1763
Vibrational Assignment
C-Cl stretch
CNC def, C-Cl stretch
C-Cl stretch
Unassigned
orthosubstituted benzene
orthosubstituted benzene
Unassigned
Unassigned
Unassigned
orthosubstituted benzene
orthosubstituted benzene
Unassigned
orthosubstituted benzene
Unassigned
Unassigned
Unassigned
orthosubstituted benzene
orthosubstituted benzene
orthosubstituted benzene
possible ketone str
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Table S5. Tentative SERS vibrational assignment for Cocaine
Raman Shift (cm-1)
418.6
526.8
617.8
695.4
735.4
793.4
847.9
914
1009
1053
1126
1178
1309
1351
1443
1489
1530
1571
1708
1768
1857
Vibrational Assignment
Aromatic Ring
Unassigned
Aromatic ring
Unassigned
Unassigned
Piperidine ring
Unassigned
Unassigned
Benzoic acid
Benzoic acid
Unassigned
Unassigned
Unassigned
Unassigned
Cyclic CH2/CH3
Unassigned
Unassigned
Aromatic ring or benzyl ring
Unassigned
Unassigned
Unassigned
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Table S6. Tentative SERS vibrational assignments for Amphetamine
Raman Shift (cm-1)
428.2
530
614.7
738.4
838.9
967.6
1009
1053
1103
1149
1195
1289
1337
1432
1484
1530
1596
Vibrational Assignment
(SO4) asym stretch
Unassigned
Aromatic ring bend
Aryl C-H wag
Alkyl c-c
(SO4)2- asym stretch
monosubstituted aromatic ring breathing mode
CC aromatic ring vibration
Unassigned
Unassigned
C-N
Unassigned
Unassigned
Unassigned
Unassigned
Unassigned
CCH aromatic ring stretch
2-
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
7000
8000
Raw data
7000
700
Wavelet filtered,
Daubechies 5,
level 3
6000
6000
5000
4
EMSC,
window=9
600
3000
3000
2
400
1
300
0
200
-1
2000
2000
1000
1000
0
4000
intensity (au)
4000
intensity (au)
intensity (au)
intensity (au)
500
5000
Autoscaled
3
0
500
1000
1500
2000
wavenumber shift (cm-1)
2500
3000
3500
0
0
500
1000
1500
2000
wavenumber shift (cm-1)
2500
3000
3500
100
0
500
1000
1500
2000
wavenumber shift (cm-1)
2500
3000
3500
-2
0
500
1000
1500
2000
2500
3000
3500
wavenumber shift (cm-1)
Figure S4. Data pre-processing of the SERS spectra. The processing is detailed below:
1. Plots of the 249 Raw Raman spectra analysed.
2. The raw Raman spectra were filtered by using Daubechies 5 wavelet function with 3 levels of decomposition. The detail coefficients
were replaced with 0s while the approximation coefficients were kept the same. The Raman spectra were then reconstructed by using
the wavelet coefficients to remove the high frequency noises and spikes (S. Mallat (1989) IEEE Pattern Analysis and Machine
Intelligence 11, 674-693).
3. The data were next normalised using extended multiplicative scatter correction using a bin size of 9 (H. Martens et al. Anal. Chem.,
2003, 75, 394-404).
4. Finally these data were then autoscaled. This is a form of scaling which mean-centres each value of the column followed by dividing
row entries of a column by the standard deviation within that column (R. Goodacre et al. (2007) Metabolomics 3, 231-241).
Principal Component 2 (11.1% TEV)
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
amphetamine
cocaine
ketamine
MDMA
MDAI
mephedrone
4
2
0
-2
-4
-6
-8
-6
-2
0
2
-4
Principal Component 1 (44.0% TEV)
4
Figure S5. Principal components analysis (PCA) scores plot of the processed SERS spectra. In general the replicate spectra are located near one
another highlighting the excellent spectral similarity. Details of PCA can be found in (R. Goodacre et al. (1998) Microbiology 144, 1157-1170).
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
PLS predictions of drugs:
The full set of results from bootstrapped PLS1 calibration (1000 iterations) for the three drugs: MDMA, MDAI and Mephedrone. For each figure
(Figures S4A-C) there are four components that are labelled as:
A. Boxplot depicting the mean training and test predictions for each sample across all 1000 bootstrapped models.. The boxes have lines at
the lower quartile, median, and upper quartile values; the whiskers are lines extending from each end of the boxes to show the extent
of the rest of the data, and outliers are marked by crosses.
B. Histogram showing the distribution of training and test predictions across all 1000 bootstrapped models.
C. Receiver operating characteristic (ROC) & ROC convex hull (ROCCH) give the true-positive (TP) rate and false positive (FP) rate of
classification; ROC curves are beneficial because they avoid having to apply a numerical threshold which defines the class boundary. A
ROC curve with an area of 1 shows 100% classification accuracy. For this study the ROC has been calculated using the full complement
of test predictions only across all 1000 bootstrapped models.
D. A confusion matrix giving the number of true positive (TP), false positive (FP), true negative (TN), false negative (FN) classifications
based upon the mean of PLS1 test predictions for each sample across the 200 bootstrapped models. This is based upon an arbitrary
classification boundary of >=0.5 (positive), <0.5 (negative). Some common metrics that can be derived from this calculation (sensitivity,
specificity, precision and accuracy) are also provided. Where:
TN
specificity =
TN + FP
precision =
TP
TP + FP
accuracy =
The above process was adapted from (K. Faulds et al. (2008) Analyst 133, 1505-1512).
TP + TN
TP + FP + TN + FN
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
A
B
PLS Predictions
2
Average PLS Predictions
40
Train predictions
Test predictions
35
1
% Frequency
PLS Prediction
30
0
-1
25
20
15
10
-2
5
-3
True positive rate
C
0
-3
Train MDMA- Test MDMA- Train MDMA+ Test MDMA+
Receiver Operating Characteristic (Test Samples)
1
ROC
ROCCH
0.8
D
MDMA+
-2
-1
0
PLS Prediction
Contingency Table
FP: 1
0.6
0.4
0
TP: 26
Sens.
Spec.
Prec.
Acc.
MDMA-
AUROC : 0.995
AUROCCH: 0.997
0.2
0
0.2
0.4
0.6
False positive rate
0.8
Figure S6A. PLS1 bootstrap results for MDMA
1
1
0.7222
0.9939
0.963
0.9453
TN: 164
FN: 10
MDMA-
MDMA+
2
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
A
PLS Predictions
Average PLS Predictions
B
1.5
35
1
True positive rate
C
25
0
% Frequency
PLS Prediction
0.5
-0.5
-1
20
15
-1.5
10
-2
5
-2.5
Train predictions
Test predictions
30
Train MDAI-
Test MDAI-
Train MDAI+
Test MDAI+
0
-3
Receiver Operating Characteristic (Test Samples)
1
ROC
ROCCH
0.8
D
0
-1
PLS Prediction
-2
Contingency Table
MDAI+
FP: 0
0.6
TP: 105
Sens. 0.9633
Spec. 1
Prec. 1
Acc. 0.9801
0.4
MDAI-
AUROC : 1
AUROCCH: 1
0.2
0
1
0
0.2
0.6
0.4
False positive rate
0.8
Figure S6B. PLS1 bootstrap results for MDAI.
1
TN: 92
FN: 4
MDAI-
MDAI+
2
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
A
B
PLS Predictions
2
1
0.5
0
30
20
-1
10
Test Mephadrone+
Train MephadroneTest MephadroneTrain Mephadrone+
True positive rate
40
-0.5
-1.5
C
Train predictions
Test predictions
50
% Frequency
PLS Prediction
1.5
Average PLS Predictions
60
Receiver Operating Characteristic (Test Samples)
1
ROC
ROCCH
0.8
0
-1.5
D
Mephedrone+
-1
-0.5
0
0.5
1
PLS Prediction
Contingency Table
FP: 0
0.6
Sens. 0.9107
Spec. 1
Prec. 1
Acc. 0.9751
0.4
Mephedrone-
AUROC : 1
AUROCCH: 1
0.2
0
TP: 51
0
0.2
0.4
0.6
False positive rate
0.8
1
Figure S6C. PLS1 bootstrap results for Mephedrone
TN: 145
FN: 5
Mephedrone-
Mephedrone+
1.5
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Figure S7A. Annotated PLS1 Loadings Plots for MDMA. The red triangles represent vibrations from MDMA, black circles represent vibrations from
MDAI and blue squares represent vibrations from Mephedrone.
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Figure S7B. Annotated PLS1 Loadings Plots for MDAI. The red triangles represent vibrations from MDMA, black circles represent vibrations from
MDAI and blue squares represent vibrations from Mephedrone.
Electronic Supplementary Material (ESI) for Analyst
This journal is © The Royal Society of Chemistry 2012
Figure S7C. Annotated PLS1 Loadings Plots for Mephedrone. The red triangles represent vibrations from MDMA, black circles represent vibrations
from MDAI and blue squares represent vibrations from Mephedrone.