Page 5 of 13
Analytical Chemistry
When δ = 10, τ decreased from 8 (standard samples) to 4
detection condition could be found in supplementary
1
2
3
4
5
6
7
8
information (Figure Sꢀ5, 6). According to the regulation
proposed before, if the δ was set at ± 10 cm−1, Sꢀ102, Sꢀ104,
Sꢀ109, Sꢀ110, Sꢀ301, Sꢀ400, Sꢀ401, Sꢀ402, Sꢀ403, Sꢀ404,
and Sꢀ501 would have 8, 22, 13, 13, 8, 10, 15, 10, 15, 10,
and 11 matching peaks with the 32 calculated Raman shifts,
respectively. If the δ was set at ± 5 cm−1, these compounds
would have 6, 14, 5, 9, 5, 6, 9, 8, 6, 4, and 5 matching peaks,
respectively. If the δ was set at ± 2 cm−1, the compounds
would have 1, 5, 1, 2, 1, 0, 1, 3, 4, 2, and 2 matching peaks,
respectively. According to these results, the δ value of 2 is
an overly strict standard that caused some samples to have
only one or even no matching peak. Hence, this setting was
not used in the subsequent study. Suppose that δ was set at ±
10 cm−1, the 11 derivatives would then contain at least 8 of
the calculated common Raman shifts. If δ was set at ± 5
cm−1, then the 11 derivatives would contain at least 4 Raman
shifts. In this study, we used a δ value of ± 10 cm−1.
(simulated samples); when δ = 5, τ decreased from 4
(standard samples) to 3 (simulated samples). These results
confirmed that setting the δ at ±10 cm−1 was appropriate for
the simulated sample. I.e., a positive result was obtained if
the sample contained at least 4ꢀ5 Raman shifts of the
calculated common shifts.
Detection of real samples by TLC-SERS. Furthermore,
the method was applied to detect real samples. Six real
samples were provided by the Shandong Institute for Food
and Drug Control and analyzed by using TLCꢀSERS. The
results were based on the 32 common calculated peaks and
the δ was set at ± 10 cm−1. Because these samples contained
at least 4ꢀ5 Raman shifts of the calculated common shifts,
they were preliminarily identified as “suspected positive
samples” for further verification in the lab. Sample 5
displayed a Raman signal at 487, 559, 736, 745, 807, 816,
916, 927, 1042, 1098, 1169, 1180, 1229, 1237, 1316, 1529,
1559, and 1584 cmꢀ1. Sample 5 contained 7 Raman shifts
that matched the calculated common shifts. This sample was
identified as a sildenafil derivative adulterant. After
comparison with the Raman shifts of 11 sildenafil
derivatives, sample 5 was found to be doped with Sꢀ109
(Figure 5). The TLCꢀSERS results were further confirmed
by using UPLCꢀQTOF/MS (Figure Sꢀ9). The result
indicated that the 32 calculated common peaks can be used
as a measure to discriminate suspected adulterants in BDSs.
Through the validation of suspected positive samples using
UPLCꢀQTOF/MS, PC strategy was found to be able to
quickly and accurately detect sildenafil derivatives in BDSs.
Generally, if the τ is fixed, the reduction of δ value will
result in an increased falseꢀnegative rate and the rise of δ
value will result in an increased false positive rate. It is
worth noting that screening or identification based on one
specific Raman peak was insufficient to produce a definitive
result; thus, identification using a series of Raman peaks (at
least four or five peaks unique to each synthetic drug) was
necessary for screening purposes11. According to the
previous results, δ should be set at ± 10 cm−1 for onꢀsite
detection. If the tested BDSs had been collected from a
relatively reliable sales channel, τ should be set at 4ꢀ5. On
the contrary, if the tested BDSs had been collected from
suspicious or unknown sales channels or had a record of
adulteration, the value of τ should be stricter to reduce the
falseꢀnegative rate. In order to set the appropriate value of τ,
we referred to other SERSꢀbased methods. In the detection
of melamine in milk samples, as long as one Raman shift at
675 cmꢀ1 was detected in a sample, that sample was
suspected to be adulterated with melamine24,25. As for the
detection of sildenafil and its derivatives, τ should be set at
2. In other words, if 2 common shifts were identified in a
sample from suspicious or unknown sales channels or with a
record of adulteration (as in the case of melamineꢀmilk), this
sample would be regarded as a suspicious sample.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Detection of simulated samples by TLC-SERS. In order
to show the feasibility of the developed TLCꢀSERS method,
10 simulated samples were separated and detected by using
the established TLCꢀSERS method. After TLC separation,
the prepared silver colloid was dropped on the separated
spot which was then visualized under 254 nm (Figure Sꢀ7).
According to the common calculated peaks, sample 19
exhibited the SERS signal at 419, 484, 553, 630, 658, 717,
904, 926, 1098, 1259, 1293, 1312, 1560, and 1583 cmꢀ1 for
the first spot. When δ was set at 10 cm−1, the results of the
detection contained 6 Raman shifts of the calculated
common shifts. When δ was set at 5 cm−1, the results
contained 3 Raman shifts of the calculated common shifts.
The SERS signal was found to be 485, 814, 909, 1004,
1109, 1025, and 1396 cmꢀ1 for the second spot. With δ set at
10 cm−1, the results of the detection contained 4 Raman
shifts of the calculated common shifts. With δ set at 5 cm−1,
results of the detection contained 3 Raman shifts of the
calculated common shifts. The SERS signal was found to be
525, 570, 733, 753, 910, 948, 1040, 1096, 1236, 1266, 1530,
and 1562 cmꢀ1 for the third spot. With δ set at 10 cm−1, the
results of the detection contained 5 Raman shifts of the
calculated common shifts. With δ set at 5 cm−1, the results of
the detection contained 4 Raman shifts of the calculated
common shifts. The SERS signal was found to be 418, 484,
535, 819, 926, 990, 1009, 1185, 1236, 1332, and 1529 cmꢀ1
for the fourth spot. With δ set at 10 cm−1, the results of the
detection contained 5 Raman shifts of the calculated
common shifts. With δ set at 5 cm−1, the results of the
detection contained 3 Raman shifts of the calculated
common shifts. The results showed that the doping substrate
was sildenafil or one of its derivatives. Furthermore, the first
spot was found to be at the same position as Sꢀ109 on the
TLC plate, suggesting that Sꢀ109 was adulterated in this
sample. The remaining spots were found to correspond to Sꢀ
301, Sꢀ400, and Sꢀ501, respectively (Figure Sꢀ8). Because
the concentration of the derivatives was very low and the
Raman enhancement was irregular on the TLC stationary
phase (silica particles), some of the Raman peaks were not
enhanced and would sometimes even disappear in the SERS
spectrum, which resulted in the loss of available commonꢀ
peak information. In addition, the differences in instrument
and sample status would inevitably lead to the fluctuation of
Raman shifts. Hence, the number of detected peaks in the
simulated samples was lower than in the standard samples.
ACS Paragon Plus Environment