J. Biomedical Science and Engineering, 2009, 2, 532-537
doi: 10.4236/jbise.2009.27077 Published Online November 2009 (http://www.SciRP.org/journal/jbise/
JBiSE
).
Published Online November 2009 in SciRes. http://www.scirp.org/journal/jbise
Application of quartz crystal nanobalance in conjunction with
a net analyte signal based method for simultaneous
determination of leucine, isoleucine and valine
Maryam Shojaei1, Abdolreza Mirmohseni2, Maryam Farbodi2
1Department of Natural sciences, Faculty of Animal sciences, University of Tabriz, Tabriz, Iran; 2Polymer Research Technology Labo-
ratory, Department of Applied Chemistry, Faculty of Chemistry, University of Tabriz, Tabriz, Iran.
Received 22 June 2009; revised 16 July 2009; accepted 19 July 2009.
ABSTRACT
The aim of the present investigation was to develop a
biosensor for the detection of amino acids, Leucine,
Isoleucine and Valine based on a quartz crystal nano-
balance. leucine (Leu), isoleucine (Ile), and valine
(Val) were selectively determined by quartz crystal
nanobalance (QCN) sensor in conjunction with net
analyte signal (NAS)-based method called HLA/GO.
An orthogonal design was applied for the formation
of calibration and prediction sets including Leu, Ile
and Val compounds. The selection of the optimal time
range involved the calculation of the net analyte sig-
nal regression plot in any considered time window for
each test sample. The searching of a region with
maximum linearity of NAS regression plot (minimum
error indicator) and minimum of PRESS value was
carried out by applying a moving window strategy.
On the base of obtained results, the differences on the
adsorption profiles in the time range between 1 and
300 s were used to determine mixtures of compounds
by HLA/GO method. The results showed that the
method was successfully applied for the determina-
tion of Leu, Ile and Val.
Keywords: Quartz Crystal Nanobalance; Net Analyte Sig-
nal; Leucine; Isoleucine; Valine; HLA/GO Method
1. INTRODUCTION
Maple Syrup Urine Disease (MSUD) is a rare autosomal
recessive metabolic disorder affecting the metabolism
of amino acids, which occurs due to a deficiency of the
activity of the mitochondrial enzyme complex bran-
ched-chain l-2-ketoacid dehydrogenase (BCKD). The
main symptom of MSUD is accumulation of the
branched chain amino acids (BCAA) leucine (Leu),
isoleucine (Ile) and valine (Val) in blood, urine and
cerebrospinal fluid. This deficiency results in mental
retardation if not detected soon after birth and l-leucine
and/or its keto acid are considered to be the main neu-
rotoxic metabolites in MSUD. So, identification and
detection of the abnormal levels of these metabolites in
urine samples are necessary to diagnosis and therapy of
these pathologies [1].
MSUD screening methods based on gas chromatog-
raphy-mass spectrometry (GC-MS) and tandem mass
spectrometry (MS-MS) have been developed. MS-MS
has been reported to be a powerful diagnostic tool in
MSUD patients. GC-MS and MS-MS have high resolu-
tions that enable them to be used to measure several
amino acids simultaneously. However, these instruments
are prohibitively expensive, and hospitals in developing
countries cannot afford it [2].
In general, the high-performance liquid chromatogra-
phy (HPLC) method has been devised to measure amino
acid levels, but developed methods have disadvantages,
which include complex sample preparation and long
analysis time [3].
Mass screening emergency for MSUD in childhood
demands to develop simple and inexpensive methods
with a rapid and quantitative response. Quartz crystal
nanobalance (QCN) is a sensing system based on the
sorption of analyte on an adsorbent material [4]. The
QCN comprises a thin vibrating AT-cut quartz wafer
sandwiched between two metal excitation electrodes.
When small amounts of mass are adsorbed at the quartz
electrode surface, the frequency of the quartz is changed
according to the well-known Sauerbrey equation [5]:
)(1026.2 2
0
6
A
m
FF    (1)
where ΔF is the measured frequency shift, F0 the origi-
nal oscillation frequency of the dry crystal, Δm the mass
change, A the piezoelectrically active area of the excita-
tion electrodes.
Due to some advantages including low cost, portabil-

M. Shojaei et al. / J. Biomedical Science and Engineering 2 (2009) 532-537 533
ity and easy on-line analysis, the quartz crystal nanobal-
ance (QCN) sensor is extensively used for the measure-
ments of mass changes in a variety of chemical and bio-
logical studies, such as determination of volatile organic
compounds [6,7], poisonous compounds [8] and immu-
noassay [9,10].
In some cases, the major drawback of the sensors
based on QCN is a lack of selectivity since along with
the analyte, other compounds usually interfere. In other
words, there is no discrimination between the sources of
the mass changes. To overcome this shortcoming, one
approach is the pattern recognition technique that can be
used for the data processing of the QCN signals for the
simultaneous determination of mixtures of compounds.
Multilinear regression (MLR), partial least squares (PLS)
and net analyte signal (NAS) are examples of multivari-
ate analytical techniques [11,12,13].
HLA/GO algorithm (Hybrid Linear Analysis pre-
sented by Goicoechea and Olivieri), one of the NAS-
based methods, has been successfully used for resolving
multicomponent mixtures. Goicoechea and Olivieri [14]
have determined tetracycline in blood serum by using
synchronous spectrofluorimetry through the HLA/GO
algorithm. This algorithm has also been applied for the
simultaneous determination of leucovorin and meth-
otrexate, by spectrophotometric [15] and Sorbic (SOR)
and benzoic (BEN) acids in fruit juice samples by using
spectroscopic signals [16]. HLA/GO has also been ap-
plied to the determination of binary mixtures of amoxy-
cillin and clavulanic acid by stopped-flow kinetic analy-
sis [17].
To our knowledge, no study reported for the detection
and determination of Leu, Ile and Val using QCN tech-
nique. In the present study, we report the simultaneous
determination of Leu, Ile and Val in the solution con-
taining some common urine analytes using polystyrene
(PS) coated QCN. NAS is utilized to process the fre-
quency data of the crystal at various times, based on
different adsorption dynamics of Leu, Ile and Val on the
PS coated QCN.
1.1. Theory Notation
An I×J data matrix R composed of the calibration re-
sponses of I samples at J times, a J×1 vector sk contain-
ing the pure adsorption profile of analyte k at unit con-
centration, and an I×1 vector ck of calibration concentra-
tions of analyte k are the used matrices and vectors
throughout the present work. The net analyte signal
(NAS) for analyte k (r∗k) is given by the following equa-
tion:
r∗k = [I − R−k(R−k)+]r = PNAS,kr (2)
where r is the adsorption profile of a given sample
(when r is the profile sk of pure k at unit concentration,
Eq.(2) becomes s∗k = PNAS,ksk), I is a J×J unitary matrix,
R−k is a J×A column space spanned by the adsorption
profile of all other analytes except k (R−k
+ is the pseudo-
inverse of R−k and A is the number of factors used to
build the model, andPNAS,k is a J×J projection matrix
which projects a given vector onto the NAS space.
The concentration of component k in an unknown sam-
ple is obtained from its adsorption profile (r) as
2
*
** )(PrPr
k
k
T
k
k
T
k
T
k
k
T
k
T
k
k
s
rs
PPss
Ps
Pss
s
c (3)
The applied method in this research involves using the
mean (uncentred) calibration profile. It is first obtained
as



I
i
cali
cal r
I
r
1
,
1 (4)
where ri,cal is the profile for the ith calibration sample.
Then the contribution of analyte k is subtracted from the
data matrix R in the following way:
calk
T
calk
kc
rc
RR
,

 (5)
where calk
c, is the mean (uncentred) calibration con-
centration of analyte k. The calculation of net sensitivity
(s*
k ) is then carried out with the following equation:









calk
T
cal
kNASk c
r
Ps
,
,
* (6)
1.2. Selection of Time Window
In the present work, the selection of the optimum range
of time window was made by calculating an error indi-
cator (EI) as a function of a moving window for each
prediction sample, using information of the NASRP
(called “net analyte signal regression plot”). NASRP is a
plot of the elements of the sample vector r*
k versus those
of s*
k and should fit a straight line through the origin,
with random residuals and slope ck. Large and correlated
residuals in this plot reveal discrepancies between the
measured profile (and thus in r*k ) and the model and,
possibly, bias in the estimated concentration. The ex-
pression for EI used in the present context is [17]:
*
2
1
2
*
22
2
4
1
r
r
sN
s
EI



















 (7)
where s is the standard deviation from the best-fitted
straight line to the NASRP (in a given adsorption region),
SciRes
Copyright © 2009 JBiSE

M. Shojaei et al. / J. Biomedical Science and Engineering 2 (2009) 532-537 534
and N is the number of points in the latter plot.
2. MATERIALS AND METHODS
2.1. Reagents and Material
All reagents used in this experiment were of analytical
grade. leucine (Leu), isoleucine (Ile), valine (Val) were
from Sigma chemicals with analytical grade. Polysty-
rene (PS) was supplied by Tabriz Petrochemical Co.,
Iran.
2.2. Instrumentation
10 MHz AT-Cut quartz crystals with gold coating on both
sides were commercially available from International
Crystal Manufacturer (ICM, Oklahoma, USA). For QCN
experiments a home made apparatus was used as de-
scribed in our previous work [18].
2.3. Procedures
A solution casting method was used to coat the polymer
over the quartz crystal electrode. Using a Hamilton mi-
cro liter syringe (Hamilton BonaduzAG, Switzerland),
4µL of PS/chloroform solution (0.3%, w/v) was dropped
on top of the gold electrode of the quartz crystal. A thin
layer of PS was obtained after solvent evaporation.
An orthogonal design was applied for the formation of
calibration and prediction sets including Leu, Ile and Val.
Orthogonal design is used in order to give the most in-
formation from the analytical system by using only a
few samples. The calibration and prediction sets were
prepared according to four-level orthogonal design.
The concentrations varied in the linear range of each
compound (50–300 mgL−1 for Leu, 100-400 mgL−1 for
Ile and Val). All solutions were filtered using a syringe
filter (0.2 µm) before injecting to the cell. Milli-Q water
was used to desorbed analyte and recover the electrode.
All measurements were carried out at room temperature
(25 °C).
3. RESULTS AND DISCUSSION
3.1. Determination of Pure Leu
The polymer-coated electrode was exposed to a constant
concentration of aqueous Leu solution (50 mgL−1).
Typical responses for Leu are shown in Figure 1. The
frequency of the crystal decreased due to the adsorption
of analyte to the surface of polymer modified electrode
according to Eq.(1) . The recorded responses showed that
the electrode is sensitive to Leu. The frequency of the
crystal was back shifted to its initial value by exposure
to the Milli-Q water indicating the full description of
analyte from the electrode surface (Figure 1).
As the concentration of analyte increased the magni-
tude of the response increased (Figure 2), the calibration
0
10
20
30
40
50
60
0200 400 600 80010001200
Time (sec)
-∆F(Hz)
c
ab
Figure 1. Typical frequency change of PS modified quartz
crystal electrode recorded upon exposure to a Leu solution
(100 mgL-1). (a) Milli-Q water, (b) Leu solution (100 mgL-1),
(c) Milli-Q.
0
50
100
150
200
250
0100 200 300 400 500 600 700 800
Time(sec)
-∆F(Hz)
d
c
b
a
Figure 2. Frequency changes of PS modified quartz crystal
electrode as a function of time exposed to various concentra-
tion of Leu solutions: (a) 50 mgL-1, (b) 100 mgL-1, (c) 200
mgL-1, (d) 300 mgL-1.
curve was constructed by plotting the frequency shifts
against the concentration of Leu (Figure 3). The re-
sponses were linear against Leu concentrations in the
range 50–300 mgL−1 and with linear regression coeffi-
cient of 0.9806 (n = 4).
3.2. Determination of Leu in the Presence of Ile
and Val
Based on the above results, the QCN sensor coated with
PS can be employed as pure Leu sensor. Since Leu, Ile
and Val are considered as the agents that exist simulta-
neously in the most urine samples of MSUD patients, it
is necessary to investigate the cross-sensitivity between
Leu, Ile and Val. So, the frequency shifts were recorded
for quartz crystal PS-coated electrode upon exposure Ile
and Val. The linear range was obtained 100–400 mgL−1
for Ile and Val. The frequency shift obtained for quartz
SciRes
Copyright © 2009 JBiSE

M. Shojaei et al. / J. Biomedical Science and Engineering 2 (2009) 532-537
SciRes Copyright © 2009
535
y = 0.7088x + 8.0678
R
2
= 0.9806
0
50
100
150
200
250
050100 150 200 250 300 350
Concentration (mgL
-1
)
-∆F(Hz)
JBiSE
Figure 3. Calibration graph for Leu solutions exposed to PS
modified quartz crystal electrode. Exposure time: 10 min.
0
50
100
150
200
0100 200300 400 500600 700 800
Time(sec)
-∆F(Hz)
a
b
c
d
Figure 4. PS modified QCN sensor response upon exposure to:
(a) Leu at the concentration of 100 mgL−1, (b) Ile at the con-
centration of 160 mgL−1, (c) Val at the concentration of 130
mgL−1, (c) mixture of Leu (100 mgL−1 )/Ile(160 mgL−1 )
/Val(130 mgL−1).
crystal PS-coated electrode upon exposure to Leu (100
mgL −1), Ile (160 mgL−1), Val (130 mgL−1) and a mixture
of Leu (100 mgL−1)/Ile (160 mgL−1)/Val (130 mgL−1)
were recorded (Figure 4). The concentrations were se-
lected from the linear range of each compound. The ob-
tained responses showed a significant change in the
shape of the frequency–time curves of Leu with includ-
ing Ile and Val (Figure 4). Then, net analyte sig-
nal-based HLA/GO method was considered to develop a
model for selectively determination of Leu, Ile and Val
compounds.
3.3. Optimization of HLA/GO Method
Selection of the optimum number of factors to be used
within the HLA/GO algorithms allows one to model the
system with the optimum amount of information. In
HLA/GO analysis of the calibration set, the PRESS
value (predicted error sum of squares) for prediction
samples varies as a function of the number of factors. In
the present work, cross-validation has been used to se-
lect the optimum number of factors for two time inter-
vals, in the range comprised between 1 and 600 s.
The selection of the optimum time region is caused to
increase the predictive ability of multivariate analysis by
discarding the non-informative parts of adsorption pro-
file from the original data.
The selection of the optimum time region for the
analysis was carried out by evaluating the best predicted
values for the prediction samples and the minimum error
EI values. In this regard, using the optimized number of
factors selected in each region, an EI was calculated for
each prediction sample, using information of the
NASRP.
0
50
100
150
200
250
300
050100 150 200 250 300
Reference Concentration (mgL-1)
Predicted HLA/GO (mgL-1)
(a)
0
50
100
150
200
250
300
350
400
100 150 200 250 300 350 400
Reference Concentration (mgL-1)
Predicted HLA/GO (mgL-1)
(b)
0
50
100
150
200
250
300
350
50100 150 200 250 300 350
Reference Concentration (mgL-1)
Predicted HLA/GO (mgL-1)
(c)
Figure 5. Predicted vs. actual concentrations for HLA/GO
calibration models (a) Leu, (b) Ile and (c) Val.

M. Shojaei et al. / J. Biomedical Science and Engineering 2 (2009) 532-537 536
Table 1. Optimization of the sensor range in the prediction of Leu, Ile and Val in the mixture by application of the NAS signal and
evaluation of the EI. (1actual concentration; 2predicted concentration)
Sample Time
Range Factor EI Leu1 Leu2ErrorEI Ile1Ile2 ErrorEI Val1 Val2 Error
1 1-300 3 0.43 60 67.7 7.7 0.04325330.65.61 0.57 300 293.1 –6.8
1-600 5 0.56 43.8 –16.50.04 346.121.1 0.57 317.1 17.1
2 1-300 3 0.42 125 120.1–5.1 0.08275251.1–13.860.06 350 365.0 15.06
1-600 5 0.36 146.021.0 0.05 142.6–22.230.77 370.9 20.9
The moving window was obtained by varying the time
range. Table 1 shows the ranges of time tested, the op-
timum number of factors for each region, the EI values
calculated, and the predicted values for Leu, Ile and Val.
The minimum EI value calculated using information of
the NASRP indicates 1-300 s as the most adequate time
region for the analysis in this case.
The optimized model was tested in the analysis of the
prediction set and plots of cpred versus cact were con-
structed (Figure 5). As it can be seen, the plot showed
very good linearity and the values of 0.9894, 0.9707 and
0.9854 were obtained as correlation coefficient for Leu,
Ile and Val, respectively.
4. CONCLUSIONS
Leucine (Leu), isoleucine (Ile), valine (Val) were simul-
taneously determined using adsorption profile data re-
corded using PS-coated QCN sensor in conjunction with
HLA/GO multivariate calibration method. Determina-
tion was based on frequency shifts of PS modified quartz
crystal electrode due to the adsorption of Leu at the sur-
face of modified electrode in the presence of Ile and Val.
The responses were linear against Leu concentrations in
the range 50-300 mgL−1 and with linear regression coef-
ficient of 0.9806 (n = 4), respectively. The selection of
optimum time ranges for each analyte separately were
performed by getting the minimum EI, based on the
minimization of the PRESS, as a function of a moving
adsorption time window. The analysis of the prediction
set was used to test the optimized model and plots of
cpred versus cact showed very good linearity. The values of
0.9894, 0.9707 and 0.9854 were obtained as correlation
coefficient for Leu, Ile and Val, respectively.
5. ACKNOWLEDGMENTS
We are most grateful the financial supports of this research project by
the University of Tabriz.
REFERENCES
[1] Barschak, A. G., Marchesan, C., Sitta, A., Deon, M.,
Giugliani, R., Wajner, M., and Vargas, C. R., (2008) Ma-
ple syrup urine disease in treated patients: Biochemical
and oxidative stress profiles, Clinical Biochemistry, 41,
317–324.
[2] Deng, C. and Deng, Y., (2003) Diagnosis of maple syrup
urine disease by determination of l-valine, l-isoleucine,
l-leucine and l-phenylalanine in neonatal blood spots by
gas chromatography-mass spectrometry, Journal of
Chromatography B, 792, 261–268.
[3] Bridi, R., Fontella, F. U., Pulrolnik, V., Braun, C. A.,
Zorzi, G. K., Coelho, D., Wajner, M., Vargas, C. R., and
Dutra-Filho, C. S., (2006) A chemically-induced acute
model of maple syrup urine disease in rats for neuro-
chemical studies, Journal of Neuroscience Methods, 155,
224–230.
[4] Lau, K. T., Micklefield, J., and Slater, J. M., (1998) The
optimization of sorption sensor array for use in ambient
conditions, Sens. Actuators. B, Chem., 50, 69–79.
[5] Sauerbrey, G. Z., (1959) The use of quartz oscillators for
weighing thin layers and for microweighing, Phys, 155,
206–222.
[6] Mirmohseni, A. and Oladegaragoze, A., (2004) Determi-
nation of chlorinated aliphatic hydrocarbons in air using
a polymer coated quartz crystal microbalance sensor
Sensors and Actuators B, 102, 261–270.
[7] Mirmohseni, A. and Oladegaragoze, A., (2003) Determi-
nation of Ammonia and Aliphatic amines in Air Using
Poly(N-vinylpyrrolidone) Coated Quartz Crystal Micro-
balance Sensors and Actuators B, 89, 164–172.
[8] Mirmohseni, A. and Alipour, A., (2002) Construction of a
sensor for the determination of cyanide in industrial ef-
fluents: A method based on Quartz Crystal Microbalance
Sensors and Actuators B, 84, 245–251.
[9] Mirmohseni, A., Shojaei, M., and Farbodi, M., (2008)
Application of a quartz crystal nanobalance to the
molecularly imprinted recognition of phenylalanine in
solution Biotechnology and bioprocess Engineering, 13,
592–597.
[10] Shojaei, M., Mirmohseni, A., and Farbodi, M., (2008)
Application of a quartz crystal nanobalance and principal
component analysis for the detection and determination
of histidine, Anal. Bioanal. Chem., 391, 2875–2880.
[11] Zhu,W., Wei,W., Nie, L., and Yao, S., (1993) Anal. Chim.
Acta, 282, 535–541.
[12] Nyberg, H., (2008) Multivariate analysis applied to a test
procedure for determining gun propelling charge weight
Part II, Partial least squares analysis, Chemometrics and
Intelligent Laboratory Systems, 92, 118–124.
[13] Mirmohseni, A., Abdollahi, H., and Rostamizadeh, K.,
(2007) Net analyte signal-based simultaneous determina-
tion of ethanol and water by quartz crystal nanobalance
sensor, Analytica Chimica Acta, 585, 179–184.
[14] Goicoechea, H. C. and Olivieri, A. C., (1999) Enhanced
synchronous spectrofluorometric determination of tetra-
cycline in blood serum by chemometric analysis. Com-
parison of partial least-squares and hybrid linear analysis
calibrations, Anal. Chem., 71, 4361–4368.
[15] Espinosa-Mansilla, A., Meras, I. D., Gomez, M. J. R.,
SciRes
Copyright © 2009 JBiSE

M. Shojaei et al. / J. Biomedical Science and Engineering 2 (2009) 532-537 537
Munoz de la Pena, A., and Salinas, F., (2002) Selection
of the wavelength range and spectrophotometric deter-
mination of leucovorin and methotrexate in human serum
by a net analyte signal based method Talanta, 58,
255–263.
[16] Marsili, N. R., Sobrero, M. S., and Goicoechea, H. C.,
(2003) Spectrophotometric determination of sorbic and
benzoic acids in fruit juices by a net analyte signal-based
method with selection of the wavelength range to avoid
non-modelled interferences, Anal. Bioanal. Chem., 376,
126–133.
[17] Munoz de la Pena, A., Espinosa-Mansilla, A., Acedo
Valenzuela, M. I., Goicoechea, H. C., and Olivieri, A. C.,
(2002) Comparative study of net analyte signal-based
methods and partial least squares for the simultaneous
determination of amoxycillin and clavulanic acid by
stopped-flow kinetic analysis, Anal. Chim. Acta, 463,
75–88.
[18] Mirmohseni, A., Milani, M., and Hassanzadeh, V., (1999)
Ion exchange properties of polypyrrole studied by elec-
trochemical quartz crystal microbalance (EQCM), Polym.
Int., 48, 873–878.
SciRes
Copyright © 2009 JBiSE