Blanchard, Alan P. (1998) Synthetic DNA Arrays; in Genetic Engineering, Vol.
20, pp. 111-123, edited by J.K. Setlow, Plenum Press, New York.
Synthetic DNA Arrays
Analytical techniques that exploit the phenomenon of sequence-specific hybridization between nucleic acids have been workhorses of modern molecular biology. With the exponentially increasing rate of genomic sequencing has come a need for hybridization techniques that can keep pace with this vast amount of information. One way is to exploit the parallelism inherent in using arrays of bound DNA as analytical tools. A convenient format for doing this is to array DNA samples, such as cDNAs or PCR products, on a flat substrate, such as nitrocellulose filters or glass plates. Arrays with several thousand samples on an area of a few square centimeters have been made and proven to be extraordinarily useful, especially in gene expression studies (1-3). But their manufacture is tedious and expensive, due mostly to the effort required to synthesize and purify the DNA samples prior to arraying them. Moreover, using naturally derived DNA samples limits one to probes that are found in nature and presents the usual handling and logistic problems when duplicating experiments in other labs. With the enormous amount of sequence information now available it is possible to consider making the probes synthetically. Over the past several years, a new set of technologies have emerged for making arrays of synthetic surface-bound oligonucleotides for doing hybridization experiments. These technologies and some of the applications for synthetic DNA arrays are the subjects of this review.
The basis of modern chemical DNA synthesis is the step-by-step coupling of activated nucleoside phosphite triesters in the proper order to produce DNA oligomers with the desired sequence (4) (Figure 1).
Note that the synthesis starts from a hydroxyl-terminated organic linker molecule that is covalently attached to a glass surface. In the usual commercial synthesis of soluble DNA oligomers, the glass surface is in the form of a fine powder of porous glass, which has a high surface area to mass ratio. By this technique, excess reagents can be washed away from the desired products at each step of the synthesis by merely filtering the mixture. The reactions proceed equally well on a flat surface (5) and in either case it is possible to use organic linkers that can be chemically cleaved, thus releasing the DNA into solution. While glass is the most common support material, many different materials can be used, such as plastics with reactive groups on the surface (6,7). Synthetic DNA is not limited to the four natural nucleosides (A,C,T and G) and a variety of chemical moieties can be introduced, such as dyes, biotin or cross-linking reagents. Some nucleoside analogues, such as 3-nitropyrrole or 5-nitroindole, when incorporated into DNA show indiscriminate base-pairing and thus serve as universal bases. While DNA synthesis is usually carried out in the 3'-5' direction (as shown), reagents exist to do 5'-3' synthesis, both with the standard acid-deprotection (Glen Research, Sterling, Virginia) or by photo-deprotection (see below) (8).
Synthetic DNA arrays can be made by either synthesizing the individual oligos in separate tubes and then attaching them, in a regular array, to a flat substrate, or the oligos can be synthesized in situ in regular arrays. The former method allows for a post-synthetic purification of the DNA before attachment but becomes impractical as the number of different oligos increases into the tens or hundreds of thousands. Such arrays can be made surprisingly dense, Yershov (9), et al., have reported making arrays on glass covered with 20 micron thick pads of polyacrylamide, each 40 microns on a side with 80 micron spacings between pads. After chemically activating the polyacrylamide, each DNA sample was delivered to its individual pad with a small pin. After completing the covalent linkage of the DNA to the polyacrylamide, the array is ready for hybridization experiments (10,11). Three-dimensional polymer pads have a capacity 100-1000 times greater than a monolayer of probe synthesized on glass, a property that makes them attractive for preparatory arrays. Livache (12) and colleagues have directed the attachment of oligonucleotides to a surface electrochemically. They prepared oligos with terminal pyrrole groups. Pyrrole can be electrochemically polymerized into an insoluble film on the surface of an electrode by applying a voltage to the electrode. A mixture of pyrrole-bearing oligos and pyrrole thus yields a film decorated with accessible oligos only on the active electrodes of an array. Different electrodes can thus be coated with different oligo bearing films by dipping the array in different oligo-pyrrole solutions while turning on selected electrodes.
Synthesizing the DNA in situ has the important engineering advantage of only requiring a small number of different reagents, no matter how many different oligos are desired on the array. The synthetic yield of the correct oligo decreases as the length of the oligo increases, thus placing a practical limit on the size of the oligos. For many applications this has proved not to be a limitation. Several strategies have been employed to synthesize different oligos in different regions of the same substrate but they all rely on either deprotecting the entire surface and exposing only selected areas to the activated phosphoramidites, or deprotecting selected regions and exposing the entire surface to the phosphoramidites.
The simplest method relies on the use of various gasket arrangements to physically isolate different regions of the array. One early experiment involved using several parallel lengths of rubber tubing regularly spaced on a flat plate, which then formed channels when pressed against the array plate (13). Different nucleoside solutions were then pumped through the channels to react only with their corresponding stripe on the array plate. This would normally result in a one-dimensional set of stripes of oligos but by rotating the channel plate by 90 degrees with respect to the array plate after each coupling cycle, it is possible to make square arrays, albeit with some restrictions on the pattern of sequences. Using this method it is possible, with only very simple apparatus, to construct, in one day, 4 copies of all 256 octapurines on a square glass plate 96 mm on a side, the spots being roughly 2 mm on a side.
A second arrangement is to use a circular cup, pressed against the array plate, to confine the synthesis reagents to a circular area (14). After each coupling the cup is displaced by, e.g., one tenth the cup diameter. This forms a series of overlapping decamers, called a scanning array, that can be used to probe RNA secondary structure or detect polymorphisms.
These mechanical techniques, while useful for illustrating the principles of array synthesis and demonstrating feasibility of various applications (13-20), are rather constrained in the size and density of the arrays they can produce. One application where this is not so important is the manufacture of primers for sequencing or PCR. If a linker molecule is used that can be cleaved by, e.g., base, the DNA can be detached from the array and used in solution. Weiler and Hoheisel (7) have made arrays of oligos on polypropylene films using a cleavable linker and found that the product from just 32 mm2 of the film is sufficient to perform a standard sequencing reaction. At that scale, 1000 primers would fit on a substrate the size of this page.
One of the most accurate, and by far the most common, methods of making high-density patterns on surfaces is photolithography. This technique, which forms the basis of all modern semiconductor ("computer chip") manufacture, works by shining light through a photolithography mask, which is much like a photographic negative, onto a light sensitive surface. The pattern on the mask, which can have features smaller than one micron, is thus reproduced on the surface. The reagents for manufacturing DNA arrays by photolithography (Figure 2) (21) are modifications of the usual phosphoramidite reagents in that the DMT group that protects the 5' hydroxyl is replaced by a protecting group that is removed by exposure to light. The synthesis proceeds by photolithographically deprotecting all the areas that are to receive a common nucleoside, coupling that nucleoside by exposing the entire array to the appropriate phosphoramidite, then after the oxidation and washing steps, repeating the procedure for the next nucleoside. To make an array of N-mers thus requires 4N cycles of deprotection and coupling, one for each of the 4 bases, times N base positions. Currently, the vast majority of synthetic oligo arrays are produced by the photodeprotection method, with features roughly 20 microns square. This photolithographic method requires 4N masks, which are specific to the pattern of sequences on the array but can be used to make many copies of the array. As the masks add considerable expense to the procedure, photolithography is best suited towards generating large numbers of identical arrays.
A more versatile, but still essentially mechanical, method for producing DNA arrays is to use the print heads out of commercial piezoelectric ink-jet printers to deliver reagents to individual spots on the array (22-25). A piezoelectric ink-jet head (Figure 3) consists of a small reservoir with an inlet port and a nozzle at the other end. One wall of the reservoir consists of a thin diaphragm with an attached piezoelectric crystal. When a voltage is applied to the crystal, it contracts laterally thus deflecting the diaphragm and ejecting a small drop of fluid from the nozzle. The reservoir then refills via capillary action through the inlet. One, and only one, drop is ejected for each voltage pulse applied to the crystal thus allowing complete control over when a drop is ejected. Such devices are inexpensive and can deliver drops with volumes of tens of picoliters at rates of thousands of drops per second. In conjunction with a computer-controlled XY stepping stage to position the array with respect to the ink-jet nozzles, it is possible to deliver different reagents to different spots on the array. Arrays of 150,000 spots can be addressed in under one minute with each spot receiving one drop of reagent (25). Ink-jets can be used in two different ways to make arrays. The nucleoside monomers and activating agent (tetrazole) can be delivered by separate ink-jets to simultaneously couple different nucleosides to different spots. The common oxidation, deprotection and washing steps can then be performed by dipping the array in the appropriate reagents. Or the deprotection reagent can be delivered to selected spots on the array which is then treated with one of the four nucleosides (24). Coupling, oxidation and wash steps are again performed uniformly across the whole array surface. The former method requires at least five ink-jets (four nucleosides plus activator) and N cycles for an array of N-mers while the latter method requires only one ink-jet, for the deprotecting agent, but 4N cycles, since each cycle couples only one of the 4 nucleosides.
Since a computer controls the pattern of reagents as the array is being made, it is as easy to make 10 arrays with different sequences as it is to make 10 identical arrays. This flexibility is perhaps the main advantage of the ink-jet approach. Achieving high density with the ink-jet approach requires one more trick. Two drops of liquid applied too closely together on a surface will tend to spread into each other and mix. For 40 picoliter drops the minimal center-to-center spacing is about 600 microns. This limits the array density achievable with the ink-jet method. One way around this is to engineer patterns in the surface chemistry of the array to produce spots of a relatively hydrophilic character surrounded by hydrophobic barriers (22,23) (Figure 4). At the small length scales involved (ca. 100 microns), surface tension is the dominant force on a drop of liquid and a hydrophobic surface will effectively prevent a drop from spreading out beyond the confines of the hydrophilic surface. There are several ways to engineer such a surface. The original suggestion was to use a laser printer to lay down a "solvent repellant grid" (22) of pigment on the array surface. Modern techniques use fluorinated alkyl silanes which covalently couple to glass and present an extremely hydrophobic surface. They can be patterned by masking the areas to remain hydrophilic, derivatizing the exposed surface with the appropriate silane and then removing the mask by dissolving it with various organic solvents. The mask itself can be formed by a photolithographic process wherein the array is covered with a thin, uniform layer of photoresist, which is then exposed to light through a photographic negative to define the array (23). The photoresist is developed, leaving behind a pattern of protective photoresist to act as a mask. Alternatively, a rubber stamp can be used to apply either the protective mask or the hydrophobic silane itself (26). Any of these methods will easily produce 100 micron diameter hydrophilic wells separated by 40 micron hydrophobic barriers, or 5000 array sites per cm2.
Hybridization to arrays built on non-porous surfaces, such as glass, can take place in volumes on the order of 2 microliters (27), which is over 1000 times less than is necessary with filters. When using oligos, the differences in melting temperature due to differing GC contents can become a problem when trying to do high-stringency experiments with large numbers of probes. This can be mitigated somewhat through the use of tetramethylammonium chloride (28), which tends to stabilize AT base pairs and bring their melting point closer to that of GC base pairs. Steric hindrance between the glass surface and the DNA can reduce hybridization yields if the linker molecule between the glass and the DNA is too short (29). Also, if the probes are too dense upon the glass, hybridization yields suffer (30). If the array is constructed with each probe on an independent electrode, hybridization times can be dramatically reduced by electrophoretically attracting the sample DNA to the probe (31). After the sample has hybridized, the stringency can be controlled by the strength of the (reversed) electric field . Better discrimination between sequences can also be obtained through engineered mismatches (32), a technique based on the observation that the difference in melting temperatures between 1 and 2 mismatches can be larger than that between 0 and 1 mismatch. 3-nitropyrrole nucleoside can be used as a mismatch to any of the 4 bases.
Detection of hybridization is generally done by labeling the sample, either radioactively, with a chemiluminescent reporter or with a fluorescent dye, then detecting the presence of the label on the array after hybridization and washing. The labeling can be done by making copies of the sample using labeled PCR primers or incorporating labeled nucleotides, or by direct chemical means (33). Dubiley and colleagues (11) have used a sort of "sandwich" assay wherein the sample is used as a template for a ligation between a 5'-phosphorylated, bound 10-mer probe and a fluorescently labeled 5-mer in solution. After the ligation, the sample was washed away and the fluorescence was imaged. Intermediate rounds of ligating 5'-phosphorylated, but unlabeled, 5-mers can improve discrimination, especially at the ends of the probes which are usually the positions least sensitive to mismatches. Since the ligation is irreversible, these arrays can only be used once.
Fluorescence can be detected with a scanning confocal microscope or a CCD camera with imaging optics, such as a microscope. Fluorescent dyes have an advantage over radioactivity in that they come in different colors so that one can distinguish between different samples in the same experiment. Radioactivity can be detected with the traditional X-ray film or phosphorimager plates, but see Eggers, et al., (34), for a discussion of very sensitive detection, using all three reporter types, with a CCD camera chip in contact with the probe. Stimpson, et al., (35) have reported an interesting real-time detection technique that involves attaching small (200 nm) particles to the sample molecules. Particles that are held close to the array surface by hybridization to the probe scatter evanescent illumination away from the surface of the array into a CCD camera. Particles in solution are generally too far away from the surface to scatter any light and are invisible. The scattered light is intense enough to allow exposures of 1/30 second, fast enough for real-time measurements of hybridization kinetics.
Once a gene has been sequenced the next important question is how that sequence varies from individual to individual, especially in individuals with a disease phenotype. Resequencing many individuals involves almost as much work, per individual, as the initial gene sequencing and is prohibitively expensive. Known mutations can be assayed by using allele-specific oligos, either as hybridization probes or PCR primers. But these methods only work for known mutations and can be expensive if the region has many known mutations, each of which must be assayed individually. Alternatively, the region in question can be amplified and labeled by PCR and hybridized against an array consisting of a sequence of overlapping oligos spanning the region. With probe oligos 20 nucleotides long, single-base mismatches are detectable and a single base mutation will affect the hybridization signal of 20 probes, giving great redundancy to the test (36). Detecting a mutation is a simple matter of noting a difference in the hybridization pattern between the reference sample and the test sample, which is conveniently done by labeling the two samples with different dyes and recording the ratio of signals from the two dyes. This two-dye method automatically controls for irregularities in the array as well as sequence-specific variations in hybridization efficiency. The position of the mutation is given directly by which oligos show a difference in hybridization intensity. A true, single-base mutation will result in a pattern of approximately 20 adjacent probes showing reduced hybridization while more complicated mutations will affect larger groups of probes. Differences in signal from isolated probes can be dismissed as artifacts. Unknown mutations consisting of isolated single-base substitutions can be identified by using a "4L" scanning array consisting of oligos corresponding to every position of the gene each augmented by 3 more oligos wherein the middle base is substituted with the three remaining bases. More complex mutations can be localized on the gene and fully characterized by sequencing just the region in question. If a list of known mutations is available, those sequences can also be included on the array, along with all their possible single-base mutations. As with any hybridization based assay, problems can arise when the sequence being scanned has repeats that are longer than the oligos being used.
With this technique, several kilobases of DNA can be scanned for mutations in a single hybridization experiment. In order to design such an array it is, of course, necessary to have the full sequence of the gene. Or viewed from a different angle, we are resequencing a piece of DNA whose sequence is almost completely known. But it is also possible to sequence unknown DNA by hybridization. By way of illustration, let's assume our unknown DNA is single stranded and only 15 nucleotides long (Figure 5). We will also assume, for the moment, that our hybridization detection only reports perfectly complementary hybridizations. Under such ideal conditions, hybridizing our unknown 15-mer against an array consisting of all 65,536 8-mers should, almost always, result in 8 perfect hybridizations, one for each of the 8 alignments of an 8-mer against a 15-mer. Given a list of the 8 perfect matches, the original sequence can be reconstructed by aligning all the 7 base overlaps. This procedure is similar to assembling the sequence of a 40,000 base cosmid given several hundred 500 base sequencing reads. It is clear how this procedure may fail; a repeated 7 base (or longer) sequence will result in an ambiguous reconstruction. Under our idealized assumptions and using an array of all 8-mers, one expects to reconstruct the correct sequence 95% of the time if the unknown sample is 180 bases or shorter. Using an array of all 1,048,576 10-mers, the procedure should work 95% of the time on samples 560 bases long, which would be comparable to the results from a standard didexoy sequencing reaction. When working with longer samples, it is best to fragment the DNA, either chemically, ultrasonically or by hydrodynamic shear through an orifice, in order to improve hybridization kinetics.
One million 10-mers is a large array, even by the photodeprotection method, so various schemes have been devised to reduce the number of probes necessary to sequence by hybridization. Bains and Smith (37) and Pevzner, et al., (38) describe arrays which have longer probes but individual spots on the array consist of well-chosen mixtures of oligos. This reduces the total number of spots on the array but it also reduces the hybridization signal since only a fraction of the molecules in a spot will have the correct sequence to participate in the hybridization. Chetverin and Kramer (39) describe a two-step procedure that first uses a large preparative array to physically sort fragments of a larger sample based on the presence of defined subsequences. The product of each spot on the preparative array is then separately eluted onto another, analytic array. Hybridization detected on a spot on the analytic array thus denotes the existence of a fragment containing both the sequence corresponding to the spot on the preparative array and the sequence of the spot on the analytic array. Large numbers, one for each spot on the preparative array, of analytical arrays are required for this procedure. See the papers by Chetverin and Kramer (39), by Ginot (40) and by Southern (41) for reviews of various aspects of oligonucleotide arrays for sequencing and mutation analysis.
With these methods of sequencing by hybridization, one uses the same array design for every DNA sample. This meshes well with the photodeprotection method, which is good at making many identical arrays. If one has the ability to easily make custom arrays, perhaps by the ink-jet method, then other strategies, which use fewer total probes, become feasible (42-44). These strategies generally start from the results of hybridizing a sample to a fixed array, e.g., all 8-mers. An attempt to reconstruct the original sequence from this data yields many possible answers consistent with the hybridization results, but it also suggests hybridization experiments that can resolve the ambiguities in reconstruction. For example, if fragments Y and Z both align with fragment X because of a repeated 7-mer in the original sequence, then additional probes designed to hybridize to XY and XZ should resolve the ambiguity (Figure 6). Of course, these probes cannot be designed until after the results of the first hybridization are available. If the second round of hybridizations doesn't suffice to reconstruct the original sequence, a third array can be designed to attempt to resolve the remaining ambiguities, and et cetera, until the full sequence is known. This process is, in theory, vastly more efficient, in terms of the total number of probes necessary for sequencing longer sequences, than any procedure based on fixed arrays.
Genetic analysis and diagnostics are two areas bound to be revolutionized by DNA arrays. For example, one of the most studied polymorphic regions is the human HLA system, which influences whether tissue from one individual can be transplanted into another individual without immunological rejection. HLA typing is a critical aspect of bone marrow transplants, which are a part of the treatment for several types of cancer, and several million people have been typed as part of the national bone marrow donor registry. Historically, tissues were typed by observing their reaction to standard antibodies, but this is a crude assay in that many different alleles may react similarly to the same antibody. Modern sequence-based typing has revealed hundreds of alleles in the HLA system (45), a number which is difficult to assay in anything other than an array format. When arrays become widespread, it will be possible to economically type potential donors with molecular precision. The same may be said of genetic diseases, for example the classical genetic diseases, sickle-cell anemia and the related B-thalassemia, have many variants with complicated interactions, indeed, some forms of B-thalassemia tend to mitigate the effects of sickle-cell anemia. Oligonucleotide arrays can provide a precise diagnosis leading to a custom-tailored treatment (46). High-density arrays have been used to test for mutations in the familial early onset breast cancer gene, BRCA1 (47). Genetic diagnostics need not be limited to the patient's gene's. Precisely identifying the genotype of an infecting bacteria or virus by array hybridization is another area that holds great promise, e.g., in classifying different strains of the enterohemorrhagic O157:H7 serogroup of E. Coli (48). This will become more important as more and more bacterial strains emerge that are resistant to common antibiotics necessitating a quick assay for the presence of resistance genes. Epidemiological studies also benefit greatly from the ability to track different strains through an infected population, a tricky process with pathogens, such as HIV, which mutate rapidly (49). A related application would be to study population dynamics of microbes in the wild. Bacteria can also be tagged by inserting a specific sequence, a "molecular barcode" (50), in the genome and the fate of many different tagged strains assayed in parallel by array hybridization.
The design of drugs will also benefit from information gleaned from array-based assays. The selection of oligonucleotides for use as anti-sense reagents is complicated by the fact that target mRNAs often have structures that don't allow much of the molecule to form hybrids with other oligonucleotides. Predicting which parts of the mRNA are accessible is difficult, but that information can be gotten empirically by testing the affinity of the target mRNA against thousands of candidate oligonucleotides on an array (51).
The parallelism inherent in array assays allows one to imagine querying the genetics of an entire genome. Instead of in depth queries of mutations in single genes, an array could be built to assay for known, informative polymorphic sites all along the genome. For example, an array of 400,000 probes, with 10 probes (to give a measure of redundancy) devoted to each polymorphic site, would be able to type an entire human genome with a resolution of less than 100KB, on average. Such an array, built with information from and used in conjunction with a BAC-end sequencing project (52) would transform the process of identifying genes associated with different phenotypes into a simple and practically deterministic procedure. Even without determining the actual gene(s) involved in important traits, such an array would be immensely useful in plant or animal breeding programs for identifying which individuals to cross.
Beyond querying the static information in DNA lies the powerful new technique of expression mapping. RNA, and in particular messenger RNA (mRNA) also specifically hybridizes to complementary DNA. If one allows that the concentration of mRNA in a cell is an indicator of the rate of protein synthesis, then determining the relative concentration of the different mRNAs reveals a highly detailed picture of the internal dynamics of a cell. Non-array based techniques (e.g., SAGE (53)) have shown the power of expression mapping in identifying genes differentially expressed between normal and cancer cells and between cells from different tumors (54). Array based techniques achieve a similar result with less work (2). One protocol for doing this involves creating a cDNA library from poly (A)+ RNA using primers incorporating a T7 RNA polymerase promoter site. The cDNA is transcribed into labeled RNA, which is then chemically sheared to improve hybridization kinetics. Applying the labeled RNA onto an array of representative sequences of the various genes and quantitating the hybridization by fluorescence measurements reveals something about the relative concentrations of the original mRNA species. Arrays made by the photodeprotection scheme have shown the ability to quantitate different mRNA species over 3 orders of magnitude with a precision better than 20%. The sensitivity of this method allows one to detect messages at the level of 1 to 3 copies per cell (55). If complications arise due to differences in the amount or density of probe in different spots, or sequence-dependent variations in hybridization efficiency, one can use the two-dye method to observe expression differences between different cell populations on a single array. Combining array-based expression analysis and panels of single-gene knockout strains of, e.g., yeast will be a powerful method of examining the interrelationships of gene function.
The accomplishments of the last decade have shown the enormous advantages of DNA array technology in providing wholesale answers to biological questions. Currently, the manufacture and use of synthetic DNA arrays is technologically intensive and it will take a fair effort to move these techniques from the demonstration stage to common laboratory practice. An infrastructure has to be built up to provide cheap and plentiful arrays along with the detection hardware and associated software to allow non-specialists to take advantage of the technology. It will no doubt be a number of years before graduate students order arrays as they currently order oligos, but it will no doubt come to pass. The biology of the past few decades has focused on individual genes and proteins, the biology of the future will have to study the complex interplay of networks of these molecules (56). DNA arrays will be a powerful tool in this endeavor.
1 Shalon, D., Smith, S.J. and Brown, P.O. (1996) Genome Research 6, 639-645.
2 DeRisi, J., Penland, L., Brown, P.O., Bittner, M.L., Meltzer, P.S., Ray, M., Chen, Y., Su, Y.A. and Trent, J.M. (1996) Nat. Genet. 14(4), 457-460.
3 Schena, M. (1996) BioEssays. 18, 427-431.
4 Gait, M.J. (ed.) (1984) Oligonucleotide synthesis, a practical approach, IRL Press, Oxford, UK.
5 Maskos, U. and Southern, E.M. (1992) Nucl. Acids Res. 20, 1679-1684.
6 Matson, R.S., Rampal, J., Pentoney, S.L., Anderson, P.D. and Coassin, P. (1995) Anal. Biochem. 224, 110-116.
7 Weiler, J. and Hoheisel, J.D. (1996) Anal. Biochem. 243(2), 218-227.
8 Pirrung, M.C., Fallon, L., Lever, D.C. and Shuey, S.W. (1996) J. Org. Chem. 61, 2129-2136.
9 Yershov, G., Barsky,V., Belgovskiy, A., Kirillov, E., Kreindlin, E., Ivanov, I., Parinov, S., Guschin, D., Drobishev, A., Dubiley, S. and Mirzabekov, A. (1996) Proc. Natl. Acad. Sci. U.S.A. 93(10), 4913-4918.
10 Khrapko, K.R., Lysov ,Y., Khorlyn, A.A., Shick, V.V., Florentiev, V.L. and Mirzabekov, A.D. (1989) FEBS Lett. 256(1-2), 118-122.
11 Dubiley, S., Kirillov, E., Lysov, Y. and Mirzabekov, A. (1997) Nucl. Acids Res. 25(12), 2259-2265.
12 Livache, T., Roget, A., Dejean, E., Barthet, C., Bidan, G., and Teoule, R. (1994) Nucl. Acids Res. 22(15), 2915-2921.
13 Southern, E.M., Maskos, U. and Elder, J.K. (1992) Genomics. 13(4), 1008-1017.
14 Southern, E.M., Case-Green, S.C., Elder, J.K., Johnson, M., Mir, K.U., Wang, L. and Williams, J.C. (1994) Nucl. Acids Res. 22(8), 1368-1373.
15 Williams, J.C. Case-Green, S.C., Mir, K.U. and Southern, E.M. (1994) Nucl. Acids Res. 22(8), 1365-1367.
16 Case-Green, S.C. and Southern, E.M. (1994) Nucl. Acids Res. 22(2), 131-136.
17 Maskos,U. and Southern, E.M. (1993) Nucl. Acids Res. 21(20), 4663-4669.
18 Maskos,U. and Southern, E.M. (1993) Nucl. Acids Res. 21(9). P 2269-2270.
19 Maskos,U. and Southern, E.M. (1993) Nucl. Acids Res. 21(9). P 2267-2268.
20 Southern, E.M. (1995) Electrophoresis 16, 1539-1542.
21 Pease, A.C., Solas, D., Sullivan, E.J., Cronin, M.T., Holmes, C.P. and Fodor, S.P. (1994) Proc. Nat. Acad. Sci. U.S.A. 91(11), 5022-5026.
22 Southern, E. (1989) PCT WO 89/10977.
23 Brennan, T.M. (1995) US Patent 5,474,796.
24 Baldeschwieler, J.D., Gamble, R.C. and Theriault, T.P. (1995) PCT WO 95/25116.
25 Blanchard, A.P., Kaiser, R.J. and Hood, L.E. (1996) Biosensors and Bioelectronics 11, 687-690.
26 Jackman, R.J., Wilbur, J.L. and Whitesides, G.M. (1995) Science 269, 664-666
27 Schena, M., Shalon, D., Davis, R.W. and Brown, P.O. (1995) Science 270, 467-470.
28 Maskos,U. and Southern, E.M. (1993) Nucl. Acids Res. (1992) 20(7), 1675-1678.
29 Guo, Z., Guilfoyle, R.A., Thiel, A.J., Wang, R., and Smith, L.M. (1994) Nucl. Acids Res. 22(24), 5456-5465
30 Shchepinov, M.S., Case-Green, S.C. and Southern, E.M. (1997) Nucl. Acids Res. 25, 1155-61.
31 Sosnowski, R.G., Tu, E., Butler, W.F., O'Connell, J.P. and Heller, M.J. (1997) Proc. Natl. Acad. Sci. 94, 1119-1123.
32 Guo, Z., Liu, Q., and Smith, L.M. (1997) Nat. Biotechnol. 15(4), 331-335.
33 Proudnikov, D. and Mirzabekov, A. (1996) Nucl. Acids Res. 24(22), 4535-4542.
34 Eggers, M., Hogan, M., Reich, R.K., Lamture, J., Ehrlich, D., Hollis, M., Kosicki, B., Powdrill, T., Beattie, K., Smith, S., et al. (1994) Biotechniques. 17(3), 516-525.
35 Stimpson, D.I., Hoijer, J.V., Hsieh, W.T., Jou, C., Gordon, j., Theriault,T., Gamble,R. and Baldeschwieler, J.D. (1995) Proc. Natl. Acad. Sci. 92, 6379-6383.
36 Chee, M., Yang, R., Hubbell, E., Berno, A., Huang, X.C., Stern, D., Winkler, J., Lockhart, D.J., Morris, M.S., and Fodor, S.P. (1996) Science 274, 610-614.
37 Bains, W. and Smith, G.C. (1988) J. Theor. Biol. 135, 303-307.
38 Pevzner, P.A., Lysov, Y., Khrapko, K.R., Belyavsky, A.V., Florentiev, V.L. and Mirzabekov, A.D. (1991) J. Biomol. Struct. Dyn . 9(2), 399-410.
39 Chetverin, A.B., and Kramer, F.R. (1994) Bio/Technology. 12, 1093-1099.
40 Ginot, F. (1997) Human Mutation 10, 1-10.
41 Southern, E.M. (1996) Trends Genet. 12, 110-115.
42 Margaritis, D. and Skiena, S. (1995) Proc. 36th IEEE Symp. Foundations of Computer Science (FOCS '95), Milwaukee WI, October 23-25, 613-620.
43 Bradley, R. and Skiena, S. (1997) First International Conf. On Computational Molecular Biology (RECOMB 97), January 20-23, 57-66.
44 Skiena, S. and Sundaram, G. (1995) J. Comp. Bio. 2, 333-353.
45 Bodmer, J.G., Marsh, S.G., Albert, E.D., Bodmer, W.F., Bontrop, R.E., Charron, D., Dupont, B., Erlich, H.A., Fauchet, R., Mach, B., Mayr, W.R., Parham, P., Sasazuki, T., Schreuder, G.M., Strominger, J.L., Svejgaard, A. and Terasaki, P.I. (1997) Tissue Antigens 49(3 Pt 2), 297-321.
46 Maggio, A., Giambona, A., Cai, S.P., Wall, J., Kan, Y.W. and Chehab, F.F. (1993) Blood 81, 239-242.
47 Hacia, J.G., Brody, L.C., Chee, M.S., Fodor, S.P., and Collins, F.S. (1996) Nat. Genet. 14(4), 441-447.
48 Caetano-Anolles, G. (1996) Nature Biotechnology 14, 1668-1674.
49 Kozal, M.J., Shah, N., Shen, N., Yang, R., Fucini, R., Merigan, T.C., Richman, D.D., Morris, D., Hubbell, E., Chee, M. and Gingeras, T.R. (1996) Nat. Med. 2(7),753-759.
50 Shoemaker, D.D., Lashkari, D.A., Morris, D., Mittmann, M. and Davis, R.W. (1996) Nat. Genet. 14(4), 450-456.
51 Milner, N., Mir, K.U. and Southern, E.M. (1997) Nat. Biotech. 15, 537-541.
52 Venter, J.C., Smith, H.O. and Hood, L. (1996) Nature 381, 364-366.
53 Velculescu, V.E., Zhang, L., Volgelstein, B. and Kinzler, K.W. (1995) Science 270, 484-487.
54 Zhang, L., Zhou, W. Velculescu, V.E., Kern, S.E., Hruban, R.H., Hamilton, S.R., Volgelstein, B. and Kinzler, K.W. (1997) Science 276, 1268-1272.
55 Lockhart, D.J., Dong, H., Byrne, M.C., Follettie, M.T., Gallo, M.V., Chee, M.S., Mittmann, M., Wang, C., Kobayashi, M.,Horton, H. and Brown, E.L. (1996) Nature Biotechnology 14, 1675-1680.
56 Blanchard, A.P. and Hood, L. (1996) Nature Biotechnology 14, 1649.