The design of a novel complementary metal oxide semiconductor
detection system for biochemical luminescence
Ude Lu
a, Ben C.-P. Hu
a, Yu-Chuan Shih
b, Chung-Yu Wu
b, Yuh-Shyong Yang
a,∗aDepartment of Biological Science and Technology, National Chiao Tung University, Institute of Biochemical Engineering, 75 Po-Ai Street, Hsinchu, Taiwan bDepartment of Electronics Engineering, National Chiao Tung University, Hsinchu, Taiwan
Received 13 May 2003; received in revised form 6 November 2003; accepted 13 November 2003
Abstract
We designed a complementary metal oxide semiconductor (CMOS) chip with accompanied accessories as a system for the detections and quantifications of biochemical luminescence. This is the first of such instruments that has been reported. The semiconductor chip was manufactured through a 0.25m CMOS standard process. A current mirror was designed in integrated circuit (IC) to amplify the signal current that was induced by chemiluminescence. Horseradish peroxidase (HRP)–luminol–H2O2system was used as an example to constitute
a useful platform for coupling to chemiluminescence reactions which produce H2O2. Glucose–glucose oxidase (GOD) reaction was coupled
with HRP–luminol–H2O2reaction to demonstrate the ability of the novel CMOS base instrument for quantifying the biological luminescence
of a variety of valuable clinical assays. Our results illustrated that the combination of the specifically designed CMOS IC and commercially available electronic devices established a simple and useful bioanalytical tool.
© 2003 Elsevier B.V. All rights reserved.
Keywords: CMOS; IC design; Bioluminescence; Enzyme chip; Biochemical analysis
1. Introduction
Following the progress of modern technology, especially in biotechnology and microelectronics, a worldwide medi-cal revolution is expected. There is a general trend toward more decentralized and immediate diagnostics (Hofmann et al., 2002; Kwakye and Baeumner, 2003; Askari et al., 2001). Thus, the development of an accurate, portable, relatively inexpensive and easy-to-use biosensor has be-come the most important issue in the healthcare industry (Baeumner et al., 2003; Choi and Gu, 2002; DeBusschere and Kovacs, 2001). The area of micro total analysis systems, also called “lab on a chip” or miniaturized analysis systems is growing rapidly (Reyes et al., 2002; Jain, 2003; Weigl et al., 2003). The current development of semiconductor chips for biosensor may overcome traditional problems and satisfy today and future’s requirements (Yang et al., 2002).
∗Corresponding author. Tel.:+886-3-5731983; fax: +886-3-5729288. E-mail address: [email protected] (Y.-S. Yang).
1.1. Luminescence assay
Bio- and chemi-luminescence are powerful tools for as-saying a variety of important biological molecules. Mod-ern electronic instruments have made it possible to mea-sure light emission precisely. Thus, ordinary chemicals or enzyme-catalyzed reactions coupled with light emitting sys-tems can be used to determine a variety of important biolog-ical molecules (Deluca, 1978; Kurittu et al., 2000; Karatani and Konaka, 2000). The horseradish peroxidase (HRP) with its substrate luminol and H2O2 reaction system is one of
the most popular chemiluminescence enzyme assay systems (Kricka and Thorpe, 1990). This reaction could be a plat-form reaction to quantify many reactions that produce H2O2
(Kricka et al., 2000; Nozaki et al., 1996; Nozaki et al., 1999). In this report, we designed the CMOS photodiodes array IC and constructed a novel instrument for the quantification of luminescence produced by biological reactions.
1.2. COMS photodiodes as chemiluminescence sensor
In the past, a photomultiplier tube (PMT) has been widely used as the sensor of a luminescence reaction. The
0956-5663/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2003.11.025
characters of cost, bulk appearance, and high power con-sumption limited its application to personalized or field used instruments. CMOS photodiode is a semiconductor light sensor that can be produced by standard industrial semiconductor procedures. Recently, the applications of charge-coupled device imager, CMOS camera and photodi-ode as array biosensors have been compared (Golden and Ligler, 2002) for biological imaging. We have demonstrated, utilizing the sophisticated HP4145 instrument, that CMOS photodiodes could apply to biochemical analysis (Lu et al., 2003). In this report, the novel design takes advantages of a standard CMOS processed IC and a common commercial multimeter to detect and quantify biological luminescence. The development of this technique is a stepping stone toward a hand held, cheap, and convenient home care instruments. This presents a good indication to the potential of the com-bination of biotechnology and IC design in microelectronics technology.
2. Materials and method
2.1. Electronic apparatus
The commercial instruments used in this report were E3646A power supply (Agilent), 34401A multimeter (Agi-lent), General Purpose Interface Bus (GPIB) card (National Instruments) and connection wire (National Instruments), and a common personal computer (PC). The whole struc-ture of the novel luminescence detection system reported in this report is shown inFig. 1.
Fig. 2. Current amplifier design diagram and parameters of CMOS chemiluminescence sensor chip. The chip was manufactured by a 0.25m CMOS standard process and worked under constant voltage of 3.3 V. The current amplifier was designed with two stages of current mirror, each of them have the factor of 103 for amplifying the current. After two-stage amplification,iD3/iD1 would be 106. Rload (10 k) was an external circuit resistant, which converted the iD3to voltage signal.
Personal computer Chemiluminescence signal
Photodiode Agilent 34401A
Resistor Current mirror
Signal current
GPIB connection
Amplified signal current
Fig. 1. The flow chart of a CMOS base luminescence detecting system. While exposing to the chemiluminescence, the CMOS photodiodes gen-erated the signal current. The signal current was amplified by the current mirrors and translated into voltage signal consequently by a resistor. Agi-lent 34401A collected the data and transferred it to the personal computer for further processing.
2.2. The design of current mirror in CMOS chip
Fig. 2 shows the electric circuit design of the CMOS chip. The mathematics model of the current mirror while the FET is working under saturation region is simply described as following (Sedra and Smith, 1998).
iD=
1 2kn
W
L(VGS− Vt)2 (1)
where iD is the current flowing through CMOS FET from
source to drain, kn a constant, W the width of the channel of CMOS FET, L the length of the channel of CMOS FET,
VGS= VG− VS, and Vtthe threshold voltage of the CMOS
One current mirror was composed of two FETs. In our study, kn, VGS, and Vt were all under the same condition,
therefore, the factor of amplification only depends on the ratio of W/L. After considering the range of original induced signal current iD1 (pico-ampere) and the range of resistor
(k), we designed the parameter of each FET as:
iD2 iD1 = W2/L2 W1/L1 = iD3 iD2 = W3/L3 W2/L2 = 25m/0.5 m 0.5 m/10 m = 10 3 (2) After two stages of amplification,iD3/iD1= 106, where iD1
is the current signal caused by chemiluminescence and iD3
the output current signal.
2.3. Chemicals and biochemicals
d-(+)-Glucose, HRP, luminol and bis–tris-propane were
purchased from Sigma (St. Louis, MO, US). H2O2 (30%
(w/w)) and H2KPO4 were purchased from Riedel-deHaën
(Buchs, Switzerland). Glucose oxidase (from Aspergillus
niger) was purchased from Fluka (Seelze, Germany).
Tris–HCl buffer was purchased from Amersham Bio-sciences (Buckinghamshire, UK). K2HPO4 was obtained
from J.T. Baker (Phillipsburg, NJ, US).
2.4. Enzyme preparation
HRP powder (1 mg or 80 U) was dissolved in 1 ml Tris–HCl buffer (0.1 M at pH 8.6). The powder of GOD (20 mg or 3000 U) was dissolved in 1 ml phosphate buffer
Fig. 3. Layout diagrams of CMOS chemiluminescence sensor chip. (a) CMOS chemiluminescence sensor chip. (b) The layout of the whole die. (c) The layout of photodiodes: the left side array (18× 150) contains two sets of the current mirror. The array on right side (18 × 173) was a reference array without current mirror. (d) The magnified diagram of current mirror and photodiodes array: each pixel of photodiode was in the size of 10m × 10 m, and two stages of current mirror amplify the current 106 times.
(0.2 M at pH 7.0). The enzyme solution was stored at
−80◦C before use. The HRP and GOD solutions were
melted in ice bath just before use and were diluted with the specified buffer. One unit of HRP is defined as 1.0 mg of purpurogallin formed from pyrogallol in 20 s at pH 6.0 and 20◦C by Sigma. One unit of GOD, defined by Fluka, will oxidize 1mol glucose/min at pH 7.0 and 25◦C.
2.5. Enzyme assay
The optimal pH value and temperature for the HRP– luminol–H2O2 system have been reported before (Thorpe and Kricka, 1986). In our experiment, the reaction mixture included luminol (1 mM), H2O2(1 mM), Tris–HCl (100 mM
at pH 8.6) and HRP (0.1–2 U) at 25◦C. An aliquot amount of HRP was first added into the cuvette, and followed by the injection of all other necessary reagents and samples. These processes were to make sure that the whole compounds were well mixed in the cuvette in a short time without extra shak-ing. The data was collected within 1 s after the final injec-tion. All enzyme assay data were the average of three mea-surements.
3. Results
3.1. The design and construction of CMOS base light detecting instrument
Figs. 1–3depicted the design of the CMOS system. The components of the system were diagramed inFig. 1.Fig. 2
was the diagram of current mirrors included in this CMOS sensor chip. The CMOS chip shown inFig. 3(a)was man-ufactured with a 0.25m CMOS standard process fabri-cated by Taiwan Semiconductor Manufacturing Company (TSMC), Hsinchu, Taiwan. The layout below the photodiode array inFig. 3(b)was not used in this report. The chemilu-minescence sensor chip was an 18× 150 photodiode array with pixel size 10m × 10 m for each photodiode. The current amplifiers were shown on the left ofFig. 3(c). The photodiode array on the right side was the reference array that did not contain the current mirror. This reference ar-ray was used to make sure that the current amplifiers make sense.
InFig. 2, resistor Rload = 10 k was an external circuit
resistant. Rloadconverts the current signal to a voltage signal.
In addition, Rload was manually changeable. This allowed
us to tune the output range of the voltage signal. As shown inFig. 2, the voltage signal was:
Vload= Vdd− Vout (3)
Current ID1(Pico Ampere)
0 50 100 150 200 250 300 350 Curre nt I D3 (Micro Ampere) 0 50 100 150 200 250 300 350
Current ID3(Pico Ampere)
0 50 100 150 200 250 300 Vout (Voltage) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 (a) (b)
Fig. 4. Simulation data of the CMOS chemiluminescence sensor chip. The electric circuit simulation tool, Simulation Program with Integrated Circuit Emphasis (SPICE), kindly offered by department of Electronics Engineering, National Chiao Tung University, Hsinchu, Taiwan, was used to simulate the circuit of a chemiluminescence sensor chip diagramed in Fig. 2. (a) ID1vs. ID3simulation data. (b) ID1vs. Vout simulation data.
The E3646A power supply supported a constant voltage
Vdd = 3.3 V to the CMOS chemiluminescence sensor
chip. The 34401A multimeter collects the Vload data and
transferred it to a personal computer (PC) through GPIB connection as shown inFig. 1for the signal flow chart.
3.2. Circuit simulation
The simulation results of the current mirrors’ (Fig. 2) am-plification effects were shown inFig. 4(a) and (b). The simu-lated relation between ID1versus ID3was shown inFig. 4(a),
and that of ID1versus Voutwas shown inFig. 4(b).
Chemi-luminescence generated current ID1 was quite linear in the
range of 30–240 pA. InFig. 4(b), the ID1–Voutcurve shows
that the Voutstayed at 0.2 V instead of linearly decreasing
when ID1was larger than 240 pA. While Vout< 0.2 V, VDS
of the FET in the second stage current mirror was too small to maintain the amplification properties.
3.3. Data processing of enzymatic reactions
The original data of Vnetload versus time was shown in Fig. 5. Vnetload was defined below as:
Vnet load = Vload− Vblank (4)
where Vload was defined inEq. (3). Vblank was the voltage
load caused by the dark current, and it was a constant 0.32 V. The integrated Vnetload data that was processed
manu-ally with Excel (Microsoft) were shown inFig. 6as typical progress curves for enzymatic reaction. The tangent slopes at the beginning (t = 0) of the curves imply the initial rate of enzymatic reactions. According to the tangent slopes at the beginning (t = 0) of the curves, the enzyme kinetics analysis could be figured out, and the relation between HRP units and initial reaction rate was shown in Fig. 7. To ob-tain the tangent slope att = 0, a function y(t) was used to
Time (sec) 0 5 10 15 20 25 Vnet load (voltage) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.1unit 0.2unit 0.4unit 0.6unit 0.8unit 1unit 1.2unit 1.6unit 2unit
Fig. 5. Change of voltage induced by the HRP–luminol–H2O2 reaction system. The HRP–luminol–H2O2 reaction condition was described in Section 2.5. The Vnetload data was collected continuously with the CMOS base luminescence detecting system.
T im e (s e c ) 0 5 1 0 1 5 2 0 Inte grate d V net loa d data (voltage) 0 4 0 8 0 1 2 0
Fig. 6. Integration of Vnetload data generated by HRP–luminol–H2O2 reaction. The integrated Vnetloaddata obtained fromFig. 5was processed manually with Excel (Microsoft).
describe each curve shown inFig. 6, and calculated the y(0) for each curve which can be differentiated at y(t),t = 0.
The Regression Wizard of software Sigma Plot 2001 was used to fit the curve shown inFig. 6. The function:
y(t) = 1+ at
b + ct (5)
was selected to fit the first 5 s of the curves shown in
Fig. 6(b), where a, b, and c are constant coefficients. Then, fromEq. (5), this equation was derived:
y(t = 0) = ab− c
b2 (6)
After the fitting, Sigma Plot 2001 returned the value of a, b, and c. The value of y(0) was calculated with the returned coefficient. The data inFigs. 7–9were applied with the same process. The R2 values for the fitting in this report were all higher than 0.998.
3.4. Kinetic data obtained by CMOS detection system
The H2O2standard curve obtained by the CMOS
chemi-luminescence sensor chip was shown in Fig. 8. A typical
Fig. 7. HRP enzyme activities profile obtained by CMOS base lumines-cence detecting system. The enzymatic activities were the initial slopes of the reaction curves inFig. 6.
H2O2 (mM)
0 2 4 6 8 10 1
initial rate (d voltage/d sec)
2 0 5 10 15 20 25 Km=1.5 ± 0.22 mM
Fig. 8. Michaelis–Menten plot of H2O2with the data obtained by CMOS base luminescence detecting system. The reaction mixture included lu-minol (1 mM), Tris–HCl (100 mM at pH 8.6), HRP (0.32 U), and H2O2 (0.1 mM to 15 mM) at room temperature. The experimental procedures were described inSection 2.5. The data process was described inFigs. 5–7 and in the text. The curve fitting and Km value were obtained with the Michaelis–Menten equation using Sigma Plot 2001.
G lu co se (m M )
0 1 0 2 0 3 0 4 0 5 0 60
Initial rate (d voltage/d sec)
0 5 1 0 1 5 2 0 2 5 Km= 3.8 ± 0.7 mM
Fig. 9. Michaelis–Menten plot of glucose with the data obtained by CMOS base luminescence detecting system. The reaction mixture for this coupled enzymes system includes 150 U GOD, 0.32 U HRP, luminol (1 mM), and Tris–HCl buffer (100 mM, pH 8.6) in final volume of 1 ml at room temperature. The mixture of luminol, Tris–HCl buffer, glucose and GOD had been incubated for 10 min at room temperature prior to the addition of HRP to start the reaction. Each data point was the average of three measurements and was processed as described inFig. 8.
Michaelis–Menten progress curve was obtained and the Km
of H2O2 was 1.5 mM as determined by a non-linear
re-gression program. The couple enzyme system using glu-cose as the variant substrate was shown inFig. 9. A typical Michaelis–Menten progress curve was also obtained and the
Kmof glucose was 3.8 mM.
4. Discussion
The current mirrors played a key role in the CMOS chemi-luminescence sensor chip. Because of the noise produced in regular electronic devices, it would be impossible for us to
detect and quantify the current signal at pico-ampere level unless the signal can be amplified without noise interfer-ence. The pico-ampere level electric signal can only be de-tected with some extremely sensitive instruments (e.g. HP 4145/4156 Semiconductor Parameter Analyzer). In this re-port, the current mirrors have been shown to faithfully am-plify the signal obtained by the CMOS sensor, and allowed us to collect the data with a common commercial multimeter.
4.1. Analysis of chemiluminescence produced from enzymatic reactions
In typical enzyme assays using spectrophotometer, the signals observed by UV-Vis absorption or fluorescence were accumulated. However, the luminescence cannot be accumulated in the chemiluminescence assays. The HRP–luminol–H2O2 reaction was a flash type
chemilumi-nescence reaction; the brightest emission happened at the beginning of the reaction as shown inFig. 5. Thus, progress curve similar to a typical enzyme assay was reconstructed (Fig. 6). The progress curves shown in Fig. 6 precisely matched the expected result for typical enzyme assays. The integrated voltage became lower when the excess enzyme was used (1.6 and 2.0 U HRP). This result was also ex-pected and defined the useful range for the enzyme assays in this report.
4.2. Comparison of the simulated and experimental data
As shown in Fig. 4(a) and (b), the responding curves were quite linear while chemiluminescence generated ID1
was located in the range of 30–240 pA. In Fig. 4(b), the
ID1–Vout curve shows that the Vout stayed at 0.2 V instead
of linearly decreasing when ID1 was larger than 240 pA.
While Vout < 0.2 V, VDS of the FET in the second stage
current mirror was too small to maintain the amplification properties.
The experimental data agreed with the result of the sim-ulation. InFig. 5, the initial Vnetload of the excess amount
of HRP (1.2–2 U) reactions were almost the same (about 2.8 V). According toEq. (4), we add the constant voltage load 0.32 V caused by the dark current to Vnetload,
Vload= Vnet load+ Vblank = 2.8 + 0.32 3.1 V (7)
According toEqs. (3) and (7), the Voutof the initial rate of
HRP (1.2–2 U) is,
Vout= Vdd− Vload= 3.3 − 3.1 = 0.2 V (8)
The result matched the phenomena in Fig. 4(b) that Vout
stayed at 0.2 V, while ID1was larger than 240 pA.
4.3. Quantification of HRP, H2O2, and glucose with a
CMOS base chemiluminescence detection system
As shown inFig. 7, the linear relation between HRP (unit) and initial rate y(0) stayed from 0.1 to 1.2 HRP units. Two
possible reasons may limit the linearly range inFig. 7. One was that the MOS FET working characters, as described in the previous section. Biochemical design may be the other reason that gave the linear range limitation. Under high HRP units, the initial rate of HRP–luminol–H2O2system would
be too fast for us to collect. At this point, the linear range of the enzyme profile was good enough for our current research. The results defined the effective assay range of this CMOS system, and helped us to set the HRP conditions in our following experiments.
The HRP–luminol system could couple with many other enzyme reactions that produce H2O2 (Kricka et al., 2000; Yang et al., 2002). Thus, our ability to determine H2O2
in-dicates that many other enzymes and biochemicals which are important for clinical diagnosis and other bio-related re-search can be determined. The relation between lumines-cence intensity and H2O2concentration is shown inFig. 8.
A typical curve that fits the Mechalis–Menten equation was obtained with our novel CMOS base instrument. The Km
value was 1.5 mM, which was in the same range with the
Km(1.1 mM) obtained from a standard PMT instrument,
Hi-tachi F4500 (Yang et al., 2002; Lu et al., 2003).
Glucose plays an important role in metabolism and is an important target for biochemical diagnosis. To determine the glucose concentration, the coupled enzyme assays, GOD and HRP, were also performed with the CMOS base instrument. The enzyme kinetic analysis of a glucose standard curve observed with the CMOS base instrument is shown inFig. 9. The Kmvalue of glucose was 3.8 mM by this method, which
was in the same range with the Km(3.4 mM) obtained from
the standard PMT instrument (data not shown).
5. Conclusion
In this study, a CMOS based chemiluminescence biosen-sor for glucose and H2O2have been reported. It established
a foundation, in equipments as well as in methods, to deter-mine various chemiludeter-minescent assays on chip. Following the pace, not only chemiluminescence but also fluorescence and UV-Vis absorption CMOS based biosensors for various important assays could also be developed.
We have demonstrated that the combination of CMOS photodiode, IC design technique and commercially avail-able electronic devices are useful to quantify biological enzyme assays. CMOS process, which is widely applied to digital and analog IC is very common and standard in the semiconductor industry. This means that CMOS based instruments have the advantages to be relatively low cost equipments for healthcare industry with mass production. In addition, CMOS showed great capability of logic pro-cessing and, in the future, allow the sensor chip to perform the data processing by itself. With some portable power supply, a new perspective to personal health care industry is expected. However, the insufficient sensitivity of CMOS photodiodes does handicap some applications. This may be
improved following the development of novel semiconduc-tor manufacture processes. Overall, the characteristics of CMOS chips make them very attractive to be developed as personalized clinical diagnostic instruments.
Acknowledgements
We thank Ms. Linyun W. Yang for proof reading the manuscript.
References
Askari, M., Alarie, J.P., Moreno-Bondi, M., Vo-Dinh, T., 2001. Applica-tion of an antibody biochip for p53 detecApplica-tion and cancer diagnosis. Biotechnol. Prog. 17, 543–552.
Baeumner, A.J., Cohen, R.N., Miksic, V., Min, J., 2003. RNA biosensor for the rapid detection of viable Escherichia coli in drinking water. Biosens. Bioelectron. 18, 405–413.
Choi, S.H., Gu, M.B., 2002. A portable toxicity biosensor using freeze-dried recombinant bioluminescent bacteria. Biosens. Bioelec-tron. 17, 433–440.
DeBusschere, B.D., Kovacs, G.T., 2001. Portable cell-based biosensor system using integrated CMOS cell-cartridges. Biosens. Bioelectron. 16, 543–556.
Deluca, M.A., 1978. Preface. Methods in Enzymology. Academic Press, New York.
Golden, J.P., Ligler, F.S., 2002. A comparison of imaging methods for use in an array biosensor. Biosens. Bioelectron. 17, 719– 725.
Jain, K.K., 2003. Nanodiagnostics: application of nanotechnology in molecular diagnostics. Expert Rev. Mol. Diagn. 3, 153–161. Karatani, H., Konaka, T., 2000. Activities of the bimodal fluorescent
protein produced by photobacterium phosphoreum strain bmFP in the luciferase reaction in vitro. Photochem. Photobiol. 71, 237– 242.
Kricka, L.J., Thorpe, G.H.G., 1990. Bioluminescent and chemiluminescent detection of horseradish peroxidase labels in ligand binder assay. In: Van Dyke, K., Van Dyke, R. (Eds.), Luminescence Immunoassay and Molecular Applications. CRC Press, Boca Raton.
Kricka, L.J., Bronstein, I., Voyta, J.C., 2000. Chemiluminescence, methods for detecting and quantitating enzyme activity. In: Ziegler, M.M., Baldwin, T.O. (Eds.), Methods in Enzymology: Bioluminescence and Chemiluminescence. Academic Press, San Diego.
Kurittu, J., Karp, M., Korpela, M., 2000. Detection of tetracyclines with luminescent bacterial strains. Luminescence 15, 291–297.
Kwakye, S., Baeumner, A., 2003. A microfluidic biosensor based on nucleic acid sequence recognition. Anal. Bioanal. Chem. 376, 1062– 1068.
Lu, U., Hu, B.C.P., Shih, Y.-C., Yang, Y.-S., Wu, C.-Y., Yuan, C.-J., Ker, M.-D., Wu, T.-K., Li, Y.-K., Hsieh, Y.-Z., Hsu, Y.-Z., Lin, C.-T., 2003. CMOS chip as luminescent sensor for biochemical reactions. IEEE Sens. J. 3, 310–316.
Nozaki, O., Iwaeda, T., Kato, Y., 1996. Amines for detection of dopamine by generation of hydrogen peroxide and peroxyoxalate chemilumines-cence. J. Biolumin. Chemilumin. 11, 309–313.
Nozaki, O., Iwaeda, T., Moriyama, H., Kato, Y., 1999. Chemiluminescent detection of catecholamines by generation of hydrogen peroxide with imidazole. Luminescence 14, 123–127.
Hofmann, O., Voirin, G., Niedermann, P., Manz, A., 2002. Three-dimensional microfluidic confinement for efficient sample delivery to biosensor surfaces: application to immunoassays on planar optical waveguides. Anal. Chem. 74, 5243–5250.
Reyes, D.R., Iossifidis, D., Auroux, P.-A., Manz, A., 2002. Micro total analysis systems. 1. Introduction, theory, and technology. Anal. Chem. 74, 2623–2636.
Sedra, A.S., Smith, K.C., 1998. Field-effect transistors (FETs). Micro-electronic Circuit. Oxford University Press, New York.
Thorpe, G.H.G., Kricka, L.J., 1986. Enhanced chemiluminescence reac-tions catalyzed by horseradish peroxidase. In: Deluca, M.A., McElroy, W.D. (Eds.), Methods in Enzymology. Academic Press, Orlando. Weigl, B.H., Bardell, R.L., Cabrera, C.R., 2003. Lab-on-a-chip for drug
development. Adv. Drug Deliv. Rev. 55, 349–377.
Yang, Y.-S., Lu, U., Hu, B.C.P., 2002. Prescription chips: toward the development of enzyme and biochemical CMOS chips. IEEE Circuits Dev. 18, 8–16.
