Advances and perspectives in aptamer arrays

Abstract

Aptamers are oligonucleotides (typically 10–60 bases in length) capable of binding target ligands with affinities similar to antibodies. The generation of high density multiplexed aptamer arrays for molecular diagnostics was first proposed nearly ten years ago for the quantification of the thousands of proteins within biological samples, including blood and urine. The tagless aptameric detection of small molecular compounds extends the application of such arrays to bioanalyses at the metabolite level. We present here a minireview on some existing technologies and highlight recent innovations that are being applied to this field, which may facilitate the vision of highly multi-parallelized arrays for the quantitative analysis of biological systems.

---

Insight, innovation, integration

Aptamers are reagents which can be developed to bind target ligands with a combination of high affinity and specificity. Given that aptamers are oligonucleotides, they are particularly suited to array based diagnostics, which can be used to directly quantify biological entities at both the proteomic and metabolomic level. The development of such devices, however, requires the coming together of technologies from the analytical, biological and computational fields. Detection and quantification of bound analytes is key to this task, particularly at the metabolomic level. We review some of the advances in this area and highlight some of the innovations which may facilitate the transfer of solution based aptamer applications to the array platform.

Introduction

Central to the fields of genomics , proteomics and metabolomics is the quantification of biological variation within living systems. Once reliably achieved this will permit the determination of the answers to many important questions; what is normal, what are the causes of diseases, how can they be detected, can they be categorized and how can they be treated? The ability to monitor the immense network of interactions within the cell requires techniques for high throughput multi-parallel bioanalysis. In this context, at the genomic level the microarray dominates, with DNA arrays utilized in a variety of tasks ranging from genetic profiling and comparative genomics to SNP (single nucleotide polymorphism) detection.¹

DNA microarray technology is, however, not without limitations. Quantitative applications are subject to many problems of measurement, possessing a limited dynamic range. Extrapolating information from the genomic level to the proteomic level is also fraught with difficulties. For instance, well documented studies show that the quantity of mRNA within a cell can relate poorly to the abundance of the corresponding protein. Protein arrays are seen as the next level in array evolution and hopes are high that array based quantification at the proteomic level can be harnessed. By protein arrays, in this instance, we refer to immobilized antibodies, tethered to a chip surface.² Bound ligands are conventionally quantified via either primary (fluorescence labeling) or secondary (matched pairs of antibodies) detection mechanisms. Protein arrays have their own flaws, and there are concerns that the level of specificity of many antibodies is not of the magnitude required for highly multiplexed arrays without a high level of cross-reactivity.

Aptamers are nucleic acid sequences (either DNA or RNA) capable of being developed to bind a range of ligand targets with affinity constants comparable to their antibody counterparts. However, they have the advantage of being smaller in size, with much greater conformational flexibility, and are believed to be capable of binding a range of target ligands that far exceeds that of antibodies; Table 1 highlights the variety of applications of aptamers. The specificity observed in aptamer binding is remarkable; they have the ability to distinguish between similar chemical entities³ and even post-translational modifications of proteins.⁴ Aptamers are conventionally derived from the process known as SELEX (systematic evolution of ligands by exponential enrichment),^5,6 in which sequences with high affinity to a target are iteratively selected and amplified via PCR from vast oligonucleotide libraries. Since the development of SELEX in 1990, researchers have been quick to see the potential of aptamers, with applications in drug candidate validation, therapeutics and diagnostics:⁷ see Fig. 1 for the number of journal articles published on aptamer sensors and diagnostics in recent years. It is easy to understand the appeal of aptameric arrays over protein arrays. The wealth of knowledge already in place in the development of oligonucleotide arrays offers a future where a researcher can potentially ‘dial up’ a chip of thousands of aptamers specific to their diagnostics needs.

Table 1 Sample of aptamers developed for various applications

Target	K d	Application
a The binding constant of this aptamer was artificially lowered for the competition assay.
Cocaine ^30	0.4–10 uM	Diagnostics
VPF/VEGF^51	0.14 nm	Developed into therapeutics for macular degeneration
Ricin-A-chain^52	7.4 nM	Detection mechanisms (anti-terror)
L-Selectin^53	2 nM	Aptamer affinity chromatography for protein purification
ERK2^4	4.7 nM	Inhibits phosphorylation for MAPK pathway analysis
PDGF-BB^54	0.65 nM^a	Used in HTS competition assays to identify small molecule drug candidates.

---

Fig. 1 Number of publications where aptamers were used for sensors and diagnostics (up to August 2008, search results from Scifinder).

Surprisingly recent years have not seen the expected explosion in aptamer array technology. However, this is an area of continuing research which is seeing the advancement of elegant solutions to existing practical difficulties. Given the remarkable diversity of applications of aptamers, there is enormous scope and desire to replicate these techniques on high density arrays, and thus similarly benefit from the near ubiquitous oligonucleotide array applications developed for aneuploidy and transcriptomics.

The aptamer revolution

The functional capabilities of aptamers are derived directly from their structure whether they are simple hairpins or complex three-dimensional G-quadruplexes,⁸ this conformational flexibility and diversity bestows the specificity which makes aptamers such powerful reagents. In the field of aptamer research the potential these molecules have as therapeutic agents is seductive (Table 1). The first aptamer based therapeutic, Macugen, went into clinical use in 2005 as a treatment for macular degeneration.⁷ Given that the standard pool of oligonucleotides used to initiate a SELEX experiment comprises around 10¹⁶ sequences which dwarfs many existing drug libraries, the attraction of oligonucleotide derived therapeutics is clear. But that is not the end of the aptamer story; aptamers are currently being employed in a variety of applications.

An exiting field in aptamer research is in the area of diagnostics, where aptamers are not only used for the detection of known biomarkers of a disease, but also in the identification of new ones. Selectivity can be engineered into aptamers through methods such as counter-SELEX³ by including rounds of negative selection against another entity during the process. SELEX methodologies have not only yielded the ability to discriminate between differences in proteins arising from post-translation modification events such as phosphorylation,⁴ but additionally reveal direct differences presented between cancerous and non-cancerous cells.⁹

Reproducing this technology using high density oligonucleotide arrays with all of their incumbent advantages would be a major breakthrough in molecular diagnostics. The synthetic nature of aptamers means they hold many advantages over antibody arrays in terms of reproducibility, reuse, amenability to tethering without loss of activity and synthesis scale. Chemical modification of aptamers, which SELEX is also suited to, widens their already immense sequence space making the scope of targets and potential specificity even greater.¹⁰ Of particular interest is the development of novel methods for tagless detection of aptamers. Methods which rely on the label-free quantification of ligands, are able to determine the concentration of small metabolites based on cooperative binding of target and reporter molecules. Techniques such as these can potentially widen the range of analytes quantifiable using aptamer arrays.

Aptamer arrays

Transposing aptamers from solution state to the microarray platform is restricted by many of the same problems associated with antibody (and to a lesser extent nucleic acid) arrays. For the detection of bound ligand, difficulties associated with multiplexing and interaction with the chip surface are problems inherent to this technology. Only comparatively recently investigations into the quantification of multiple protein targets on a single array were reported, and required solution optimization for characterizing multiple proteins in a single environment.^11,12 This has long been known to be a problem in antibody arrays where conditions suitable for the hybridization of one protein lead to precipitation of another. Aptamers are short oligonucleotides and so are intrinsically better suited to the microarray format, as binding assays, albeit to complementary sequences, are commonplace in transcriptomics.

Cross-reactivity becomes problematic with an increasing number of reagents on a multiplexed antibody array. This is particularly true in sandwiched antibody arrays, where matched pairs of antibodies are employed for the detection of ligand binding. Although aptamers can be derived with high levels of specificity to single proteins, can this level of specificity be maintained across an entire proteome? In one study with as few as four reagents on a single array, non-specific interactions occurred between one of the aptamers and other targets, although this was a truncated version of the optimized aptamer.¹³ Such complications are well known in nucleic acid arrays but levels of non-specificity can be monitored and predicted.¹⁴

Ligand quantification

The time and effort spent on developing aptamers that bind targets with sub-nanomolar affinities would be wasted without an equally sensitive method of detection. Fortunately, due to the widespread use and availability of DNA arrays, a multitude of techniques have been developed which are readily amenable to aptamer arrays. Depending upon the detection mechanism the surface must be either treated or constructed from specific materials, and the immobilization methods must position the aptamer and target such that binding triggers the appropriate signal or response. A low, non-specific and background signal must be observed and the positioning of the DNA must not inhibit or alter the strength of the interactions between target and probe. The method of attachment to the surface also warrants consideration: there already exist detailed reviews into the mechanisms for attaching DNA onto surfaces¹⁵ and this shall not be covered here. Worth highlighting again is the process of using linker molecules to tether DNA to surfaces; if a decision is taken to use a spacer molecule, the chemical composition of the spacer as well as its length have both been shown to affect the signal.¹⁶

A variety of detection mechanisms have been applied to aptamer arrays including: electrochemical,^17–19surface plasmon resonance imaging (SPRi),^20–23 acoustic,²⁴ and various optical methods,^11,25,26 each with varying degrees of success. Often the thrombin binding aptamer is used to validate these technologies as its behavior and characteristics are well studied and the aptamer has previously proved versatile enough to transfer to numerous surface chemistries. Thrombin also has the advantage of being a short (as few as 15 bases in length) DNA sequence, which can be synthesized with high efficiency and also helps to keep the running costs low.

Without doubt fluorescence is the most widely used and developed method for detection, with a multitude of assay variations utilizing both direct and indirect labeling. The direct labeling of target proteins obviously adds cost and sample preparation stages. Cho et al. have recently demonstrated how arrays capable of screening against multiple targets can achieve detection at pM levels.¹¹ Complications can arise with the use of labels, either by altering the protein chemistry or from interactions between the tags and the target aptamers, which can lead to either high background noise or a decrease in aptamer specificity. A substantial body of research exists showing the interaction of DNA and π-rich compounds like those typically used in fluorescence labeling.²⁷ For the high throughput screening of large numbers of metabolites or proteins, the ideal situation would be to use fluorescence on “dirty” samples that have not been labeled and subjected to little or no purification, as seen with homogeneous extraction methods.

Béra Abérem et al.have shown the detection of unlabeled targets with pM sensitivity using a polythiophene derivative that undergoes changes to both conformational and optical properties in the presence of single-stranded nucleic acids.²⁸ The method works by mixing polythiophene complexes with equal quantities of the chosen ssDNAaptamer, the resultant DNA–polymer duplex can be spotted onto any surface and is then ready for use. It is important to note that whilst the target itself is not labeled with a dye, the addition of a label to the aptamer itself is essential. When present in the duplex form the fluorescence of the label is quenched, upon addition of the target protein the aptamer undergoes a structural change that weakens the interaction between itself and the polymer. This allows the attached fluorophore to fluoresce. Future uses of the technique might be dependent upon the chosen aptamer undergoing conformational changes that will significantly alter the DNA–polymer interaction, but this method has been shown to have a 62 pM limit of detection of unlabeled thrombin.

Fluorescence methods require arrays to have no complex surface modifications or complicated technology, and as such there may be a limited financial driving force to making them reusable. Additionally current commercial scanners can resolve to 1 μm, and are capable are of scanning large chips areas within minutes; as such, fluorescence will remain the bench mark for competing technologies.

As target molecules get smaller the use of large labels may have to be reconsidered, and this will be particularly true for the quantification of small molecular metabolites.²⁹ FRET based techniques that detect the conformational rearrangement arising from ligand binding have great potential for the detection of small metabolites. As too do the “modular aptameric sensors” developed by the Stojanovic lab which have been shown to produce excellent sensitivity.³⁰ These aptamers consist of a “signaling domain” in addition to a “recognition domain”; the cooperative metamorphosis of the aptamer upon metabolite binding facilitates the binding of a “reporter” dye (malachite green), Fig 2. This has been shown to be selective to a range of targets down to the nanomolar range: ATP, flavin mono-nucleotide (FMN), and theophylline. Recent work by Kumar and co-workers at the Institute for Biological Resources and Functions in Japan have demonstrated a technique known as the “analyte-dependent oligonucleotide modulation assay”, ADONMA.^31,32 This method splits the aptamers into two non-functional halves, one of which is immobilized onto a chip’s surface, the second is labeled with a fluorophore and added to solution. Upon addition of a specific target both halves combine and the fluorescence is localized onto the array’s surface. Examples such as these and many similar variants further demonstrate the flexibility and ruggedness of aptamer detection mechanisms.

Fig. 2 Schematic and structure of the FMN sensor. The modular aptameric sensor is composed of two parts. The first section is capable of binding flavin mono-nucleotide (FMN), and termed the FMNaptamer (FMNA). Having bound FMN, the second section is capable of binding malachite green (MG), via the MG aptamer (MGA), which produces enhanced fluorescence for detection.

Alternative technologies that allow real-time and tagless detection are available but can often require a more complex surface and technology platform increasing assay costs. Li et al. at the University California Irvine have recently displayed the power of SPRi in conjugation with RNA arrays. They describe a novel method for immobilizing RNA onto their arrays using a T4 RNA ligation reaction, which generates a surface density of immobilized RNA on each spot as high as 4.0 × 10¹² molecules per cm².²² When SPRi is coupled with an enzymatic amplification using horseradish peroxidase and tetramethylbenzidene, HRP-TMB, the sensitivity can be improved by a factor of 10 000, allowing the detection level of thrombin to reach the sub-picomolar range. As demonstrated by Wang et al. with their study on human immunoglobulin E aptamers (IgE), SPRi can be easily combined with microfluidic devices,²⁰ allowing for high throughput screening whilst maintaining sensitivity comparable to that of SPR spectrometry and quartz crystal microbalances.

Electrochemical techniques have similar experimental and labeling processes as fluorescence, in that target molecules and proteins can be tagged directly with redox active compounds, which produce currents when they come into contact with the electrode’s surface. Ultimately the power of this platform comes from the tagless detection route. There has been an explosion in recent years in the field of carbon nanotubes, a full review on their use in detecting biomolecules can be found elsewhere.¹⁸ The study of carbon nanotube field effect transistors, CNT-FETs, in their various guises shows that they are ideal as label-free biosensors .^19,33–35 CNT-FETs are highly sensitive to their immediate surrounding solution environments on scales comparable to the Debye length. Due to this sensitivity of the immediate environment, Maehashi et al. demonstrated with the detection of IgE how the smaller size of aptamers, when compared to antibodies, aids and increases the detection limits.³⁴ An alternative route to the electrochemical detection mechanism is electrochemical impedance spectroscopy (EIS),^36,37 Du et al. from the Changchun Institute of Applied Chemistry, have recently pushed this technique using thrombin and ATPaptamers, to the sub-nanomolar detection limit.³⁸ Electrochemical devices are seen, amongst some, as a possible cost effective alternative to optical methods. Previously there was a shortfall in matching the limits of detection seen in fluorescence,³⁹ but the technology and sensitivity have seen progression and amenability to micro-fabricated devices such as arrays may see further advances in detection sensitivity.

The future? The new techniques for label-free detection are pushing sensitivities that are capable of detecting biologically relevant concentrations. When analyzing “dirty” samples, proving the identity and purity of the captured species could prove difficult. This is where mass spectrometry could provide the answer.⁴⁰ The power of mass spectrometry techniques and the information that can be gained is well known and barely needs discussing. Now the commercial sector is exploiting the benefits with new products such as the iPLEX Gold assay from Sequenom, using matrix assisted laser desorption/ionization with time of flight mass spectrometry (MALDI-TOF MS) technology on their more recent SNP chips.⁴¹ Recent work has shown the power and applicability of using MALDI-TOF MS on DNA modified surfaces⁴² for detecting aptamers.⁴³ Laboratories will need to balance throughput, detail of analysis and equipment running costs, but evident from the literature is that there is a technology with the capability and price to cater for all.

The outlook for aptamer arrays

Critical to the future of aptamer arrays is the systematic development of aptamers to varied ligand targets. In 1999 it was predicted that in response to the requirement to raise markers for the entire human proteome, and with the advance of high throughput robotic SELEX methodologies, it would be feasible to generate 10⁴aptamers per year.¹⁰ Even a generous assessment of the articles currently in press reveals this prediction has fallen well short of the mark. Undoubtedly the sensitivity of the private sector to these molecules restricts the access of the general scientific public to all available data, and so determining the true figure would be pure guesswork. It should be noted though that SELEX is still a highly specialized process and is not as popular amongst the wider scientific community as would be expected. In the absence of a “journal of failed experiments” (although one of the authors (WR) wishes there were one) it is difficult to ascertain whether there is any generic reason whether certain ligands may be unsuitable as aptamer targets. To date there remains only one reported case of a failed SELEX experiment.⁴⁴

When considering the role of aptamer arrays for the analysis of complex biological samples, blood and serum, any problems arising from cross-reactivity between targets and aptamers have to be scaled against the limitations of current analytical technologies.³⁵ As more than one aptamer can be developed to a single target,^6,45 combinatorial analysis of these sequences with suitable statistical methodologies (plus controls) should partially alleviate these problems.¹⁰ On-chip optimization of aptamer sequences and binding conditions has also proved to be beneficial.¹¹ Evolutionary multi-objective optimization should be particularly well suited to this task; a technique which has already been applied with some success to increase the range of metabolites detectable by GC-TOF-MS in biological samples.^46,47 Using this procedure many objectives can be optimized simultaneously through an iterative process—moving the aptamers to points of no-coincidence within the interaction space. While this process may prove expensive in terms of the number of chips required to “tune” the aptamers, replicating the process in silico may prove a more efficient option. ‘On-chip’ studies into the ligand binding profile of aptamers have been able to interrogate large portions of the sequence landscape.⁴⁸ From these data quantitative structure activity (QSAR) models similar to those used in the development of small molecular drug candidates⁴⁹ can be derived which can act as surrogates for the real systems. The plasticity offered by aptamers in this way is something unavailable with antibodies.

This highlights the true power of oligonucleotide arrays in the analysis of aptamerprotein interactions. SELEX is commonly used in the study of transcription factor binding sites. Using the best sequences selected from the SELEX process, weight matrices are constructed and used to search the genome for putative binding sites. However, the trained models (weight matrices) often correlate poorly with the observed dissociation constants, owing to the training set being based purely on good binders, with no information on those sequences with medium and low affinity.⁵⁰ Microarrays are adept for studying the sequence binding profiles of DNA–protein interactions, due to the magnitude of sequences that can be analyzed on a single chip. Utilizing high density arrays, the sequence space of aptamer/transcription factor binding sites can be interrogated in an unbiased manner, recording information from a wider spectrum of sequence affinities. It has been proposed that the reduction of feature sizes on oligonucleotide arrays will now make it feasible to map the entire sequence space of an aptamer up to ten bases in length.⁴⁸ The resultant sequences derived from array based assays can be used to train a variety of statistical models as outlined in Fig. 3. These models have applications not only in transcriptomics and the design of aptamer arrays, but also in the engineering of drug specificity, sensors and many other applications. This of course relies on a suitable detection mechanism which will not influence the model; many of the tagless techniques discussed within this review would be particularly amenable to the task.

Fig. 3 Generation of QSAR models using high density oligonucleotide arrays to explore sequence space. Results from SELEX experiments can be used to seed array experiments which explore the structural features of the aptamers.⁵⁵ This can be performed iteratively with the generation of models which are validated and improved using experimental data.

Improved detection is the key in the development of large multiplexed aptamer arrays and the advantages of label-free target quantification methods are transparent. SomaLogic is a company which specializes in the development of aptamer arrays. The photo-crosslinking technology utilized in the capturing of bound protein to aptamer reagents is by now possibly the most well proven in this area, having already been applied to the quantification of seventeen protein targets within a ‘dirty’ sample.¹² Whilst the technology may be limited when applied to small analytes, advances in the quantification of these targets are ongoing. Given the commercial opportunities and recent innovations in surface based analysis techniques, large scale molecular diagnostic arrays may not be too far away.

References

1. P. Jares, Ultrastruct. Pathol., 2006, 30, 209–219.
2. G. MacBeath, Nat. Genet., 2002, 32, 526–532.
3. R. D. Jenison, S. C. Gill, A. Pardi and B. Polisky, Science, 1994, 263, 1425–1429.
4. S. D. Seiwert, T. Stines Nahreini, S. Aigner, N. G. Ahn and O. C. Uhlenbeck, Chem. Biol., 2000, 7, 833–843.
5. A. D. Ellington and J. W. Szostak, Nature, 1990, 346, 818–822.
6. C. Tuerk and L. Gold, Science, 1990, 249, 505–510.
7. D. H. J. Bunka and P. G. Stockley, Nat. Rev. Microbiol., 2006, 4, 588–596.
8. R. F. Macaya, P. Schultze, F. W. Smith, J. A. Roe and J. Feigon, Proc. Natl. Acad. Sci. U. S. A., 1993, 90, 3745–3749.
9. M. Blank, T. Weinschenk, M. Priemer and H. J. Schlusener, J. Biol. Chem., 2001 DOI:10.1074/jbc.M100347200.
10. E. N. Brody, M. C. Willis, J. D. Smith, S. Jayasena, D. Zichi and L. Gold, Mol. Diagn., 1999, 4, 381–388.
11. E. J. Cho, J. R. Collett, A. E. Szafranska and A. D. Ellington, Anal. Chim. Acta, 2006, 564, 82–90.
12. C. Bock, M. Coleman, B. Collins, J. Davis, G. Foulds, L. Gold, C. Greef, J. Heil, J. S. Heilig, B. Hicke, M. Nelson Hurst, G. M. Husar, D. Miller, R. Ostroff, H. Petach, D. Schneider, B. Vant-Hull, S. Waugh, A. Weiss, S. K. Wilcox and D. Zichi, Proteomics, 2004, 4, 609–618.
13. T. G. McCauley, N. Hamaguchi and M. Stanton, Anal. Biochem., 2003, 319, 244–250.
14. O. Greiner and P. J. R. Day, Anal. Biochem., 2004, 324, 197–203.
15. S. Balamurugan, A. Obubuafo, S. Soper and D. Spivak, Anal. Bioanal. Chem., 2008, 390, 1009–1021.
16. S. Balamurugan, A. Obubuafo, S. A. Soper, R. L. McCarley and D. A. Spivak, Langmuir, 2006, 22, 6446–6453.
17. T. G. Drummond, M. G. Hill and J. K. Barton, Nat. Biotechnol., 2003, 21, 1192–1199.
18. S. N. Kim, J. F. Rusling and F. Papadimitrakopoulos, Adv. Mater. (Weinheim, Ger.), 2007, 19, 3214–3228.
19. H. Yoon, J.-H. Kim, N. Lee, B.-G. Kim and J. Jang, ChemBioChem, 2008, 9, 634–641.
20. Z. Wang, T. Wilkop, D. Xu, Y. Dong, G. Ma and Q. Cheng, Anal. Bioanal. Chem., 2007, 389, 819–825.
21. C. Corne, J.-B. Fiche, D. Gasparutto, V. Cunin, E. Suraniti, A. Buhot, J. Fuchs, R. Calemczuk, T. Livache and A. Favier, Analyst, 2008, 133, 1036–1045.
22. Y. Li, H. J. Lee and R. M. Corn, Nucleic Acids Res., 2006, 34, 6416–6424.
23. Y. Li, H. J. Lee and R. M. Corn, Anal. Chem., 2007, 79, 1082–1088.
24. M. D. Schlensog, T. M. A. Gronewold, M. Tewes, M. Famulok and E. Quandt, Sens. Actuators, B, 2004, 101, 308–315.
25. Y. Wang, D. Li, W. Ren, Z. Liu, S. Dong and E. Wang, Chem. Commun., 2008, 2520–2522.
26. C. Lin, Y. Liu and H. Yan, Nano Lett., 2007, 7, 507–512.
27. B. A. Armitage, DNA Binders and Related Subjects, Top. Curr. Chem., 2005, 253, 55–76.
28. M. Béra Abérem, A. Najari, H.-A. Ho, J.-F. Gravel, P. Nobert, D. Boudreau and M. Leclerc, Adv. Mater. (Weinheim, Ger.), 2006, 18, 2703–2707.
29. N. de-los-Santos-Álvarez, M. a. J. Lobo-Castañón, A. J. Miranda-Ordieres and P. Tuñón-Blanco, TrAC, Trends Anal. Chem., 2008, 27, 437–446.
30. M. N. Stojanovic, P. de Prada and D. W. Landry, J. Am. Chem. Soc., 2001, 123, 4928–4931.
31. R. Yamamoto-Fujita and P. K. R. Kumar, Anal. Chem., 2005, 77, 5460–5466.
32. S. C. B. Gopinath, T. S. Misono and P. K. R. Kumar, Crit. Rev. Anal. Chem., 2008, 38, 34–47.
33. K. Besteman, J. O. Lee, F. G. M. Wiertz, H. A. Heering and C. Dekker, Nano Lett., 2003, 3, 727–730.
34. K. Maehashi, T. Katsura, K. Kerman, Y. Takamura, K. Matsumoto and E. Tamiya, Anal. Chem., 2007, 79, 782–787.
35. W. B. Dunn, N. J. C. Bailey and H. E. Johnson, Analyst, 2005, 130, 606–625.
36. F. Lisdat and D. Schäfer, Anal. Bioanal. Chem., 2008, 391, 1555–1567.
37. D. Xu, D. Xu, X. Yu, Z. Liu, W. He and Z. Ma, Anal. Chem., 2005, 77, 5107–5113.
38. Y. Du, B. Li, H. Wei, Y. Wang and E. Wang, Anal. Chem., 2008, 80, 5110–5117.
39. R. J. Hamers, Nature, 2008, 454, 708–709.
40. M. Mrksich, ACS Nano, 2008, 2, 7–18.
41. J. Perkel, Nat. Methods, 2008, 5, 447–453.
42. H. Tsubery and M. Mrksich, Langmuir, 2008, 24, 5433–5438.
43. L. W. Dick and L. B. McGown, Anal. Chem., 2004, 76, 3037–3041.
44. K. K. H. Stanlis and J. R. McIntosh, J. Histochem. Cytochem., 2003, 51, 797–808.
45. D. M. Tasset, M. F. Kubik and W. Steiner, J. Mol. Biol., 1997, 272, 688–698.
46. S. O'Hagan, W. B. Dunn, M. Brown, J. D. Knowles and D. B. Kell, Anal. Chem., 2005, 77, 290–303.
47. S. O’Hagan, W. B. Dunn, J. D. Knowles, D. Broadhurst, R. Williams, J. J. Ashworth, M. Cameron and D. B. Kell, Anal. Chem., 2007, 79, 464–476.
48. C. L. Warren, N. C. S. Kratochvil, K. E. Hauschild, S. Foister, M. L. Brezinski, P. B. Dervan, G. N. Phillips and A. Z. Ansari, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 867–872.
49. Chemoinformatics in Drug Discovery, ed. T. I. Oprea, Wiley-VCH, Weinheim, 2004.
50. M. Djordjevic, Biomol. Eng., 2007, 24, 179–189.
51. L. S. Green, D. Jellinek, C. Bell, L. A. Beebe, B. D. Feistner, S. C. Gill, F. M. Jucker and N. Janjic, Chem. Biol., 1995, 2, 683–695.
52. J. R. Hesselberth, D. Miller, J. Robertus and A. D. Ellington, J. Biol. Chem., 2000, 275, 4937–4942.
53. T. S. Romig, C. Bell and D. W. Drolet, J. Chromatogr., B: Biomed. Sci. Appl., 1999, 731, 275–284.
54. L. S. Green, C. Bell and N. Janjic, Biotechniques, 2001, 30, 1094–1111.
55. M. Platt, W. Rowe, J. Knowles, P. J. Day and D. B. Kell, Integr. Biol., 2009 10.1039/b814892a.

---