Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Signal Oscillation Is Another Reason for Variability in Microarray-Based Gene Expression Quantification

Abstract

Microarrays have been widely used for various biological applications, such as, gene expression profiling, determination of SNPs, and disease profiling. However, quantification and analysis of microarray data have been a challenge. Previously, by taking into account translational and rotational diffusion of the target DNA, we have shown that the rate of hybridization depends on its size. Here, by mathematical modeling of surface diffusion of transcript, we show that the dynamics of hybridization on DNA microarray surface is inherently oscillatory and the amplitude of oscillation depends on fluid velocity. We found that high fluid velocity enhances the signal without affecting the background, and reduces the oscillation, thereby reducing likelihood of inter- and intra-experiment variability. We further show that a strong probe reduces dependence of signal-to-noise ratio on probe strength, decreasing inter-microarray variability. On the other hand, weaker probes are required for SNP detection. Therefore, we recommend high fluid velocity and strong probes for all microarray applications except determination of SNPs. For SNP detection, we recommend high fluid velocity with weak probe on the spot. We also recommend a surface with high adsorption and desorption rates of transcripts.

Citation: Singh R (2013) Signal Oscillation Is Another Reason for Variability in Microarray-Based Gene Expression Quantification. PLoS ONE 8(1): e54753. https://doi.org/10.1371/journal.pone.0054753

Editor: Andrew R. Dalby, University of Westminster, United Kingdom

Received: October 14, 2012; Accepted: December 17, 2012; Published: January 21, 2013

Copyright: © 2013 Raghvendra Singh. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This work was supported by the Indian Institute of Technology Kanpur grant IITK/CHE/20090282 to RS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The author has declared that no competing interests exist.

Introduction

Although all cells of a multicellular organism share the same genes, their expression depends on the cell type and their microenvironment, which is responsible for difference in their phenotype and functionality [1]. Further, genes are switched on and off during various stages of development as well as in routine physiological functions such as cell-cycle [2], circadian clock [3], DNA repair [4], stress response [5], growth [6], and differentiation [7]. Besides normal variations in mRNA level, it has been found that gene expression switching is responsible for various pathological conditions including cancer [8]. For these reasons, DNA microarray is an important diagnostic tool, which has been used for determination of single nucleotide polymorphism [9], [10], gene expression profiling [11], profiling of diseases such as: diabetes [12], [13], HIV [14], [15], Alzheimer's disease [16], [17], Parkinson's disease [18], [19], Lyme disease [20], cancer [21]–[24], and malaria [25]. However, major challenge lies in understanding and analysis of thousands of noisy data points as well as in reducing the inter-experiment and inter-microarray variability [26]–[29]. The variability of microarray data has been attributed to biological variations, sample labeling error, technical variability, probe design, cross hybridization, low abundance of transcripts, and variability associated with control genes [28], [30]–[33]. To improve inter- and intra- microarray reproducibility, various design and statistical methods have been given, which have improved quantification and analysis of data obtained from these arrays [34]–[38]. Besides improving the data analysis, other aspects of DNA hybridization, such as its rate and efficiency, have been predicted using reaction-diffusion models. Previously, by taking into account rotational and translational diffusion, we have shown that the size of transcript plays an important role in the rate of hybridization [39]. Other models of surface DNA hybridization have also been given. For example, a model has been suggested that takes into account nonspecific adsorption of the target DNA followed by its 2D diffusion, which enhanced the hybridization efficiency [40]. Furthermore, Gadgil et. al. have predicted the minimum inter spot distance required to improve detection sensitivity using a reaction diffusion model [41].

Here, we modeled diffusion of a transcript in a microchannel over a spot and show that the hybridization process is inherently oscillatory. Since probe DNAs, the small complementary molecules (∼25 nucleotides) [42] immobilized on the surface, will be adsorbed inside and not accessible to transcripts in the solution, we hypothesize that solution-phase hybridization may be negligible in comparison to surface hybridization. Thus, we assume that the transcript adsorbs on the surface nonspecifically [40], then, diffuses on the site to find the probe DNA. However, desorption rate constant is very low at the spot, where it gets hybridized to the specific probe, than at other places on the surface. We further assume that hybridization reaction is much faster than the surface adsorption and diffusion of the target DNA and, thus, can be neglected. Our results show that signal oscillation is damped and increasing fluid velocity not only enhances its intensity but also reduces the amplitude of oscillation. We further found that although high fluid velocity reduces oscillation of the background signal as well, it does not increase its intensity. Thus, our study identifies fluid velocity as an important operational factor to reduce intra- and inter experiment variability. Furthermore, we show that for a strong probe the signal-to-noise ratio does not depend on the probe strength. Thus, immobilization of strong probes on the spot will improve inter-microarray reproducibility. In contrast, a weaker probe may be required for SNP detection.

Mathematical Model

The diffusion of target DNA in the solution (Fig. 1A) is given by vector equation:with initial condition: and boundary conditions:

at x  =  0

at x  =  L

at y  =  0

at y  =  L

at z  =  h

at z  =  0 and at the spot

at z = 0 and excluding the spot

Diffusion of target DNA at the surface (at z = 0) is given by vector equation:

At the spot,and at everywhere on the surface except the spot,

with initial condition: and boundary conditions: at z  =  0 and all edges of the surface,

where,

Velocity vector:

Due to channel size, the flow is considered laminar. Thus,

Vy  = 0 and Vz  = 0

Average concentration of the hybridized DNA has been calculated by integral:and average background intensity by

The signal to noise ratio at a particular hybridization time is calculated as:Where, P: concentration of the target DNA in hybridization channel (mol/m3); P0: concentration of the target DNA in sample (mol/m3); (Psavg)specific: average concentration of hybridized target DNA on the spot (mol/m2); (Psavg)background: average concentration of nonspecifically bound target DNA on the surface (mol/m2); D: solution diffusivity of target DNA (m2/s); : velocity vector (m/s); Ps: concentration of DNA on the surface (mol/m2); Ds: diffusivity of DNA on the surface (mol/m2); θ0: concentration of the adsorption sites (mol/m2); ka: adsorption rate constant at the surface (m3/s/mol); kd1: desorption rate constant at the spot (s−1); kd2: desorption rate constant at the surface excluding the spot (s−1); Vxmax: mid-point fluid velocity in x-direction (m/s); As: area of the spot (m2); Ass: area of the surface excluding the spot (m2); : unit surface normal vector.

Values of range of surface adsorption, desorption rate constants, solution and surface diffusivities, representative concentration of target DNA in the sample, and DNA adsorption sites on the surface have been taken from literature for DNA hybridization [43], [44], [45]. The equations 1-9 have been solved using COMSOL Multiphysics 4.3a.

thumbnail
Download:
Figure 1. Schematic of hybridization channel and variation of hybridization intensity on the spot.

(A) A square channel of side length L  = 100 μm and depth h  = 10 μm has been shown. At the mid-point of the channel, there is a spot of radius R = 10 μm on which target DNA hybridizes. The fluid flows in x-direction with velocity Vx. (B) Pictures of spot intensity (xy-plane) at indicated time from COMSOL Multiphysics for a transcript concentration (P0) of 0.5 μM in the sample, adsorption rate constant (ka) of 0.1 m3/mol/s, specific desorption rate constant (kd1) of 1×10−8 s−1, concentration of adsorption sites (θ0) of 0.000104 mol/m2, surface diffusivity (Ds) of 2×10−13 m2/s, solution diffusivity of 2×10−11 m2/s, nonspecific desorption rate constant (kd2) of 0.3 s−1 and the mid-point velocity (Vxmax) of 5 mm/s. The color coded intensity is in mol/m2.

https://doi.org/10.1371/journal.pone.0054753.g001

Results

Higher fluid velocity increases hybridization intensity and reduces signal oscillation

Transcript in a microchannel (Fig. 1A) undergoes diffusion in solution and reaches the surface, where it is adsorbed nonspecifically. Following adsorption, it diffuses on the surface to reach the spot where probe molecules are immobilized and hybridizes. Concurrent to 2D diffusion and hybridization, DNA molecules desorb from the surface back to the solution, creating a feedback loop. The mathematical modeling of these reactions shows that concentration of DNA hybridized on the spot (Fig. 1B) as well as that of nonspecifically bound on the surface oscillates as hybridization progresses. The amplitude decreases, resulting in dampening and decreasing the variability of signal with time. For a fluid velocity of Vxmax  = 0.5 mm/s, signal reaches a peak at t  = 12.4 h, then decreases sharply to reach close to the background level at t  = 80.9 h (Fig. 2A). Similar to the specific signal, damped oscillation of low amplitude is also seen in the background intensity. Increasing fluid velocity not only reduces the oscillation but also increases signal to noise ratio (Fig. 2A–E), improving microarray resolution. Although amplitude of oscillation continuously decreases with increase in fluid velocity, the signal to noise ratio saturates at Vxmax of 10 mm/s (Fig. 2E), suggesting that there may be an optimum velocity for DNA hybridization.

thumbnail
Download:
Figure 2. Higher fluid velocity increases hybridization intensity and reduces signal oscillation.

Computations were performed at the indicated mid-point velocity (Vxmax) for a transcript concentration (P0) of 0.5 μM in the sample, adsorption rate constant (ka) of 0.1 m3/mol/s, specific desorption rate constant (kd1) of 1×10−8 s−1, concentration of adsorption sites (θ0) of 0.000104 mol/m2, surface diffusivity (Ds) of 2×10−13 m2/s, solution diffusivity of 2×10−11 m2/s, and nonspecific desorption rate constant (kd2) of 0.3 s−1. Signal-to-noise ratio was calculated at a hybridization time of 24 h.

https://doi.org/10.1371/journal.pone.0054753.g002

Higher concentration of transcript does not improve signal resolution

Since quantification of scantily expressed genes has been a challenge, we studied the effect of transcript concentration on signal resolution. Although increasing transcript concentration increases the signal linearly, it has no effect on signal-to-noise ratio (Fig. 3C). Further, the oscillation is also not affected by the target DNA concentration (Fig. 3A, B), implying that sample concentration of the transcript does not improve signal resolution and reproducibility.

thumbnail
Download:
Figure 3. Higher concentration of transcript does not improve signal resolution.

Computations were performed for indicated concentrations of the transcript in the sample for adsorption rate constant (ka) of 0.1 m3/mol/s, specific desorption rate constant (kd1) of 1×10−8 s−1, concentration of adsorption sites (θ0) of 0.000104 mol/m2, surface diffusivity (Ds) of 2×10−13 m2/s, solution diffusivity of 2×10−11 m2/s, nonspecific desorption rate constant (kd2) of 0.3 s−1, and the mid-point velocity (Vxmax) of (A) 1000 mm/s (B) 1 mm/s and (C) signal intensity and signal-to-noise ratio for the mid-point velocity (Vxmax) of 1 mm/s at the indicated hybridization time.

https://doi.org/10.1371/journal.pone.0054753.g003

Lower surface diffusivity of transcript enhances the signal while it simultaneously increases the amplitude of oscillation

DNA microarrays have been traditionally used to compare gene expression under different treatment conditions although their application in quantifying absolute concentration of different transcripts, including splice variants, is equally significant, specifically in disease profiling, and has been increasing progressively. Since one of the major reactions in hybridization process is surface diffusion of the transcript, its 2D diffusivity, which decreases as the transcript length increases, is an important parameter that governs the signal intensity. We found that decreasing surface diffusivity of the transcript for a fixed solution diffusivity increases both signal-to-noise ratio (Fig. 4B) and amplitude of oscillation (Fig. 4A), suggesting that although signal resolution becomes better at any time-point, its time-variability increases with increase in transcript length, which may lower the inter-experiment reproducibility.

thumbnail
Download:
Figure 4. Lower surface diffusivity of transcript enhances the signal while it simultaneously increases the amplitude of oscillation.

Computations were performed for indicated ratio of surface to solution diffusivity (RD) for a transcript concentration (P0) of 0.5 μM in the sample, adsorption rate constant (ka) of 0.1 m3/mol/s, specific desorption rate constant (kd1) of 1×10−8 s−1, concentration of adsorption sites (θ0) of 0.000104 mol/m2, solution diffusivity of 2×10−11 m2/s, nonspecific desorption rate constant (kd2) of 0.3 s−1, and the mid-point velocity (Vxmax) of 1 mm/s. Signal-to-noise ratio was calculated at a hybridization time of 24 h.

https://doi.org/10.1371/journal.pone.0054753.g004

A strong probe lowers the inter-microarray variability while a weak probe may be required for SNP detection

Probe selection, including its specificity and affinity to the transcript, constitutes an integral part of microarray design. Since specific desorption rate constant depends on probe length and its GC content, we studied its effect on hybridization. When specific desorption rate constant is low due to a strong probe immobilized on the spot, increasing it by four orders of magnitude has no effect on either the dynamics of signal or signal-to-noise ratio (Fig. 5A, B, C), implying that selection of a strong probe causes signal to be independent of probe design, improving inter-microarray reproducibility. On the other hand, if a weak probe, which means high specific desorption rate constant, has been immobilized on the surface, increasing the desorption rate constant decreases signal to noise ratio sharply (Fig. 5C), suggesting that a weak probe may increase inter-microarray variability although it may be preferred for SNP detection.

thumbnail
Download:
Figure 5. A strong probe lowers the inter-microarray variability while a weak probe may be required for SNP detection.

Computations were performed for indicated ratios of specific/nonspecific desorption rate constants (kd1/kd2) for a transcript concentration (P0) of 0.5 μM, the adsorption rate constant (ka) of 0.1 m3/mol/s, nonspecific desorption rate constant (kd2) of 0.3 s−1, concentration of adsorption sites (θ0) of 0.000104 mol/m2, solution diffusivity of 2×10−11 m2/s, surface diffusivity (Ds) of 2×10−13 m2/s, and the mid-point velocity (Vxmax) of (A) 1000 mm/s (B) 1 mm/s and (C) Signal-to-noise ratio was calculated for Vxmax  = 1 mm/s at a hybridization time of 24 h.

https://doi.org/10.1371/journal.pone.0054753.g005

Specific and nonspecific equilibrium constants have opposite effects on signal resolution while both increase the oscillation

Adsorption and desorption rate constants are measures of how fast a transcript adsorbs and releases from a surface, respectively, while equilibrium constant is a measure of amount of surface adsorbed DNA, which remains in equilibrium with that in the solution. Thus, specific equilibrium constant is a measure of transcript hybridized on the spot while the nonspecific equilibrium constant is a measure of transcript adsorbed nonspecifically on the surface. Along with probe selection, surface characteristics play an important role in the hybridization process. We found that increasing the specific equilibrium constant while keeping the nonspecific equilibrium constant fixed increases signal intensity without affecting the background, thus, enhancing the resolution (Fig. 6A, B). In contrast, increasing nonspecific equilibrium constant increases both the signal and the background, decreasing signal-to-noise ratio (Fig. 6C, D). Further, increasing either the specific or nonspecific equilibrium constant increases amplitude of oscillation, implying that higher equilibrium constants may increase inter-experiment variability (Fig. 6A, C).

thumbnail
Download:
Figure 6. Specific and nonspecific equilibrium constants have opposite effects on signal resolution while both increase the oscillation.

Computations were performed for indicated specific (Kspec, m3/mol) and nonspecific (Knonspec, m3/mol) equilibrium constants for a transcript concentration (P0) of 0.5 μM, concentration of adsorption sites (θ0) of 0.000104 mol/m2, solution diffusivity of 2×10−11, surface diffusivity (Ds) of 2×10−13 m2/s, and the mid-point velocity (Vxmax) of 1 mm/s and (A) for a specific desorption rate constant (kd1) of 1×10−8 s−1 (B) Signal-to-noise ratio was calculated at a hybridization time of 24 h and for a specific desorption rate constant (kd1) of 1×10−8 s−1 (C) for a nonspecific desorption rate constant (kd2) of 0.3 s−1(D) Signal-to-noise ratio was calculated at a hybridization time of 24 h and for a nonspecific desorption rate constant (kd2) of 0.3 s−1.

https://doi.org/10.1371/journal.pone.0054753.g006

Discussion

We found that dynamics of surface DNA hybridization is oscillatory, which is a characteristics of a feedback mechanism [46], [47]. In this case, feedback mechanism is 2D diffusion and desorption of transcript from surface back to the solution. Slower the surface diffusion and/or the desorption rate, higher the delay in feedback, and higher the amplitude of oscillation. Increasing fluid velocity causes mixing in the solution, increasing convective mass transfer, thereby, enhancing the signal and reducing the amplitude of oscillation. Although fluid velocity causes mixing effect, the overall hybridization rate is governed by slower 2D diffusion and adsorption processes. Thus, after a certain velocity signal-to-noise ratio saturates. While high signal-to-noise ratio is preferred for detection in a single experiment, signal oscillation may be responsible for inter-experiment and inter-microarray variability, along with other reasons [48]. Thus, hybridization at higher fluid velocity may increase sensitivity and reproducibility of microarray, as have been found experimentally [49], [50]. In agreement with our study, time series data obtained using microarray showed genome wide oscillation [51], [52] although it is not clear whether a part of it may be attributed to that inherent with microarrays.

Increasing transcript concentration increases the signal regardless of the fluid velocity. Since background signal also increases, transcript concentration has no effect on signal to noise ratio. Thus, a minimum concentration of target DNA is required so that the signal can be detected by available methods.

Lower surface diffusivity of transcript improves the signal as it reduces the likelihood of DNA diffusing out of the spot. However, it also increases the amplitude of oscillation since the feedback mechanism from the surface becomes even slower. Thus, a surface on which diffusivity of transcript is low increases the signal to noise ratio, improving detection in a single experiment. On the other hand, due to higher oscillation, inter-experiment and inter-microarray variability may increase.

Desorption rate constant of transcript at the spot depends on probe design. A strong probe of longer length and/or high GC content will decrease the specific desorption rate constant. The study shows that for such probes the signal-to-noise ratio does not depend on the probe design, thereby, reducing the inter-microarray variability. On the other hand, for SNP detection, it is desired that increasing desorption rate constant due to single nucleotide mismatch lowers the signal to noise ratio, differentiating the transcripts with polymorphism. Therefore, for SNP detection, a probe of shorter length and lower GC content may be preferred in agreement with recent experimental finding [53].

Increasing surface adsorption rate constant while keeping the nonspecific equilibrium constant unchanged improves the signal-to-noise ratio since it increases the specific binding at the spot without affecting the background. On the other hand, increasing the nonspecific equilibrium constant while keeping the specific equilibrium constant unchanged decreases the signal-to-noise ratio due to increased background intensity. Since adsorption and nonspecific desorption are surface properties, it is desirable to have a surface that has high adsorption and desorption rates of DNA so that specific binding is high while background is low, improving resolution in a specific experiment. However, higher adsorption rate aggravates the delay in feedback by the 2D diffusion followed by desorption, increasing the signal oscillation. Thus, low adsorption rate on the surface will improve inter-experiment and inter-microarray reproducibility.

In summary, a number of design and operational implications can be drawn from the study (Table 1). First, vigorous and uniform mixing, either in a microfluidic device or through agitation, is desired. Second, although sensitivity of microarray continues to be a major challenge [28], transcript concentration beyond the sensitivity threshold may not reduce microarray variability. Further, strong probe will improve signal-to-noise ratio and inter-microarray reproducibility. Although higher GC content is preferred, it may be primarily controlled by annealing temperature requirement of the probes. On the other hand, probes of longer length can be used although their effect on signal specificity needs further investigation [28]. In the range of target DNA concentration considered in the study, signal intensity at a given time increases linearly with increase in transcript concentration. Thus, microarray can be calibrated to measure the absolute concentration. However due to oscillation, either the hybridization time should be fixed or the experiment should be performed under vigorous mixing for absolute concentration measurement. Furthermore, lower surface-to-solution diffusivity ratio of target DNA improves signal-to-noise ratio. Since 2D diffusivity depends on polymer length much strongly than the solution diffusivity [54], transcripts of longer length may improve resolution. Although for measurement of ratios of the same transcript under different treatment conditions its length may not be important, for comparison of amounts of different transcripts including splice variants the signal intensity should be corrected for difference in transcript length.

thumbnail
Download:
Table 1. Effect of operational and design factors on microarray variability.

https://doi.org/10.1371/journal.pone.0054753.t001

Author Contributions

Conceived and designed the experiments: RS. Performed the experiments: RS. Analyzed the data: RS. Contributed reagents/materials/analysis tools: RS. Wrote the paper: RS.

References

  1. 1. Majumdar MK, Thiede MA, Mosca JD, Moorman M, Gerson SL (1998) Phenotypic and functional comparison of cultures of marrow-derived mesenchymal stem cells (MSCs) and stromal cells. J Cell Physiol 176: 57–66.
  2. 2. Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, et al. (1998) Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization. Mol Biol Cell 9: 3273–3297.
  3. 3. Doyle MR, Davis SJ, Bastow RM, McWatters HG, Kozma-Bognar L, et al. (2002) The ELF4 gene controls circadian rhythms and flowering time in Arabidopsis thaliana. Nature 419: 74–77.
  4. 4. Amundson SA, Patterson A, Do KT, Fornace AJ Jr (2002) A nucleotide excision repair master-switch: p53 regulated coordinate induction of global genomic repair genes. Cancer Biol Ther 1: 145–149.
  5. 5. Yan SF, Fujita T, Lu J, Okada K, Shan Zou Y, et al. (2000) Egr-1, a master switch coordinating upregulation of divergent gene families underlying ischemic stress. Nat Med 6: 1355–1361.
  6. 6. Richards JS (1994) Hormonal control of gene expression in the ovary. Endocr Rev 15: 725–751.
  7. 7. Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, et al. (1999) Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc Natl Acad Sci U S A 96: 2907–2912.
  8. 8. Miki Y, Swensen J, Shattuck-Eidens D, Futreal PA, Harshman K, et al. (1994) A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Science 266: 66–71.
  9. 9. Gunderson KL, Steemers FJ, Lee G, Mendoza LG, Chee MS (2005) A genome-wide scalable SNP genotyping assay using microarray technology. Nat Genet 37: 549–554.
  10. 10. Keating BJ, Tischfield S, Murray SS, Bhangale T, Price TS, et al. (2008) Concept, design and implementation of a cardiovascular gene-centric 50 k SNP array for large-scale genomic association studies. PLoS One 3: e3583.
  11. 11. Duggan DJ, Bittner M, Chen Y, Meltzer P, Trent JM (1999) Expression profiling using cDNA microarrays. Nat Genet 21: 10–14.
  12. 12. Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, et al. (2003) PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 34: 267–273.
  13. 13. Jailwala P, Waukau J, Glisic S, Jana S, Ehlenbach S, et al. (2009) Apoptosis of CD4+ CD25(high) T cells in type 1 diabetes may be partially mediated by IL-2 deprivation. PLoS One 4: e6527.
  14. 14. Geiss GK, Bumgarner RE, An MC, Agy MB, van 't Wout AB, et al. (2000) Large-scale monitoring of host cell gene expression during HIV-1 infection using cDNA microarrays. Virology 266: 8–16.
  15. 15. Meyers JH, Justement JS, Hallahan CW, Blair ET, Sun YA, et al. (2007) Impact of HIV on cell survival and antiviral activity of plasmacytoid dendritic cells. PLoS One 2: e458.
  16. 16. Blalock EM, Geddes JW, Chen KC, Porter NM, Markesbery WR, et al. (2004) Incipient Alzheimer's disease: microarray correlation analyses reveal major transcriptional and tumor suppressor responses. Proc Natl Acad Sci U S A 101: 2173–2178.
  17. 17. Nunez-Iglesias J, Liu CC, Morgan TE, Finch CE, Zhou XJ (2010) Joint genome-wide profiling of miRNA and mRNA expression in Alzheimer's disease cortex reveals altered miRNA regulation. PLoS One 5: e8898.
  18. 18. Grunblatt E, Mandel S, Maor G, Youdim MB (2001) Gene expression analysis in N-methyl-4-phenyl-1,2,3,6-tetrahydropyridine mice model of Parkinson's disease using cDNA microarray: effect of R-apomorphine. J Neurochem 78: 1–12.
  19. 19. Shehadeh LA, Yu K, Wang L, Guevara A, Singer C, et al. (2010) SRRM2, a potential blood biomarker revealing high alternative splicing in Parkinson's disease. PLoS One 5: e9104.
  20. 20. Revel AT, Talaat AM, Norgard MV (2002) DNA microarray analysis of differential gene expression in Borrelia burgdorferi, the Lyme disease spirochete. Proc Natl Acad Sci U S A 99: 1562–1567.
  21. 21. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, et al. (1999) Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 286: 531–537.
  22. 22. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, et al. (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415: 530–536.
  23. 23. Bittner M, Meltzer P, Chen Y, Jiang Y, Seftor E, et al. (2000) Molecular classification of cutaneous malignant melanoma by gene expression profiling. Nature 406: 536–540.
  24. 24. Lodes MJ, Caraballo M, Suciu D, Munro S, Kumar A, et al. (2009) Detection of cancer with serum miRNAs on an oligonucleotide microarray. PLoS One 4: e6229.
  25. 25. Le Roch KG, Zhou Y, Blair PL, Grainger M, Moch JK, et al. (2003) Discovery of gene function by expression profiling of the malaria parasite life cycle. Science 301: 1503–1508.
  26. 26. Spruill SE, Lu J, Hardy S, Weir B (2002) Assessing sources of variability in microarray gene expression data. Biotechniques 33: 916-920, 922–913.
    • 27. Lee ML, Kuo FC, Whitmore GA, Sklar J (2000) Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridizations. Proc Natl Acad Sci U S A 97: 9834–9839.
    • 28. Draghici S, Khatri P, Eklund AC, Szallasi Z (2006) Reliability and reproducibility issues in DNA microarray measurements. Trends Genet 22: 101–109.
    • 29. Sato F, Tsuchiya S, Terasawa K, Tsujimoto G (2009) Intra-platform repeatability and inter-platform comparability of microRNA microarray technology. PLoS One 4: e5540.
    • 30. Novak JP, Sladek R, Hudson TJ (2002) Characterization of variability in large-scale gene expression data: implications for study design. Genomics 79: 104–113.
    • 31. Budhraja V, Spitznagel E, Schaiff WT, Sadovsky Y (2003) Incorporation of gene-specific variability improves expression analysis using high-density DNA microarrays. BMC Biol 1: 1.
    • 32. Lee PD, Sladek R, Greenwood CM, Hudson TJ (2002) Control genes and variability: absence of ubiquitous reference transcripts in diverse mammalian expression studies. Genome Res 12: 292–297.
    • 33. de Jonge HJ, Fehrmann RS, de Bont ES, Hofstra RM, Gerbens F, et al. (2007) Evidence based selection of housekeeping genes. PLoS One 2: e898.
    • 34. Newton MA, Kendziorski CM, Richmond CS, Blattner FR, Tsui KW (2001) On differential variability of expression ratios: improving statistical inference about gene expression changes from microarray data. J Comput Biol 8: 37–52.
    • 35. Workman C, Jensen LJ, Jarmer H, Berka R, Gautier L, et al. (2002) A new non-linear normalization method for reducing variability in DNA microarray experiments. Genome Biol 3: research0048.
    • 36. Lawrence ND, Milo M, Niranjan M, Rashbass P, Soullier S (2004) Reducing the variability in cDNA microarray image processing by Bayesian inference. Bioinformatics 20: 518–526.
    • 37. Han T, Melvin CD, Shi L, Branham WS, Moland CL, et al. (2006) Improvement in the reproducibility and accuracy of DNA microarray quantification by optimizing hybridization conditions. BMC Bioinformatics 7 Suppl 2S17.
    • 38. Wang X, Ghosh S, Guo SW (2001) Quantitative quality control in microarray image processing and data acquisition. Nucleic Acids Res 29: E75–75.
    • 39. Singh R, Nitsche J, Andreadis ST (2009) An integrated reaction-transport model for DNA surface hybridization: implications for DNA microarrays. Ann Biomed Eng 37: 255–269.
    • 40. Chan V, Graves DJ, McKenzie SE (1995) The biophysics of DNA hybridization with immobilized oligonucleotide probes. Biophys J 69: 2243–2255.
    • 41. Gadgil C, Yeckel A, Derby JJ, Hu WS (2004) A diffusion-reaction model for DNA microarray assays. J Biotechnol 114: 31–45.
    • 42. Dalma-Weiszhausz DD, Warrington J, Tanimoto EY, Miyada CG (2006) The affymetrix GeneChip platform: an overview. Methods Enzymol 410: 3–28.
    • 43. Dandy DS, Wu P, Grainger DW (2007) Array feature size influences nucleic acid surface capture in DNA microarrays. Proceedings of the National Academy of Sciences of the United States of America 104: 8223–8228.
    • 44. Chan V, Graves DJ, Fortina P, McKenzie SE (1997) Adsorption and surface diffusion of DNA oligonucleotides at liquid/solid interfaces (vol 13, pg 320, 1997). Langmuir 13: 5514–5514.
    • 45. Lukacs GL, Haggie P, Seksek O, Lechardeur D, Freedman N, et al. (2000) Size-dependent DNA mobility in cytoplasm and nucleus. J Biol Chem 275: 1625–1629.
    • 46. Harootunian AT, Kao JP, Paranjape S, Tsien RY (1991) Generation of calcium oscillations in fibroblasts by positive feedback between calcium and IP3. Science 251: 75–78.
    • 47. Ueda HR, Hagiwara M, Kitano H (2001) Robust oscillations within the interlocked feedback model of Drosophila circadian rhythm. J Theor Biol 210: 401–406.
    • 48. Bryant PA, Smyth GK, Robins-Browne R, Curtis N (2011) Technical variability is greater than biological variability in a microarray experiment but both are outweighed by changes induced by stimulation. PLoS One 6: e19556.
    • 49. Schaupp CJ, Jiang G, Myers TG, Wilson MA (2005) Active mixing during hybridization improves the accuracy and reproducibility of microarray results. Biotechniques 38: 117–119.
    • 50. Adey NB, Lei M, Howard MT, Jensen JD, Mayo DA, et al. (2002) Gains in sensitivity with a device that mixes microarray hybridization solution in a 25-mu m-thick chamber. Analytical Chemistry 74: 6413–6417.
    • 51. Ptitsyn AA, Zvonic S, Gimble JM (2007) Digital signal processing reveals circadian baseline oscillation in majority of mammalian genes. PLoS Comput Biol 3: e120.
    • 52. Klevecz RR (2000) Dynamic architecture of the yeast cell cycle uncovered by wavelet decomposition of expression microarray data. Funct Integr Genomics 1: 186–192.
    • 53. Tan JC, Patel JJ, Tan A, Blain JC, Albert TJ, et al. (2009) Optimizing comparative genomic hybridization probes for genotyping and SNP detection in Plasmodium falciparum. Genomics 93: 543–550.
    • 54. Wong JSS, Hong LA, Bae SC, Granick S (2011) Polymer Surface Diffusion in the Dilute Limit. Macromolecules 44: 3073–3076.