A DNA-based pattern classifier with in vitro learning and associative recall for genomic characterization and biosensing without explicit sequence knowledge
© Lee et al.; licensee BioMed Central Ltd. 2014
Received: 5 April 2014
Accepted: 18 October 2014
Published: 6 November 2014
Genetic material extracted from in situ microbial communities has high promise as an indicator of biological system status. However, the challenge is to access genomic information from all organisms at the population or community scale to monitor the biosystem’s state. Hence, there is a need for a better diagnostic tool that provides a holistic view of a biosystem’s genomic status. Here, we introduce an in vitro methodology for genomic pattern classification of biological samples that taps large amounts of genetic information from all genes present and uses that information to detect changes in genomic patterns and classify them.
We developed a biosensing protocol, termed Biological Memory, that has in vitro computational capabilities to “learn” and “store” genomic sequence information directly from genomic samples without knowledge of their explicit sequences, and that discovers differences in vitro between previously unknown inputs and learned memory molecules. The Memory protocol was designed and optimized based upon (1) common in vitro recombinant DNA operations using 20-base random probes, including polymerization, nuclease digestion, and magnetic bead separation, to capture a snapshot of the genomic state of a biological sample as a DNA memory and (2) the thermal stability of DNA duplexes between new input and the memory to detect similarities and differences. For efficient read out, a microarray was used as an output method. When the microarray-based Memory protocol was implemented to test its capability and sensitivity using genomic DNA from two model bacterial strains, i.e., Escherichia coli K12 and Bacillus subtilis, results indicate that the Memory protocol can “learn” input DNA, “recall” similar DNA, differentiate between dissimilar DNA, and detect relatively small concentration differences in samples.
This study demonstrated not only the in vitro information processing capabilities of DNA, but also its promise as a genomic pattern classifier that could access information from all organisms in a biological system without explicit genomic information. The Memory protocol has high potential for many applications, including in situ biomonitoring of ecosystems, screening for diseases, biosensing of pathological features in water and food supplies, and non-biological information processing of memory devices, among many.
Nucleic acid technology has become an indispensible tool in medical diagnosis, microbial ecology, environmental microbiology, etc. by providing specific, sensitive detection of genes in chemically and biologically complex backgrounds. However, genome-enabled studies have focused on specific individual organisms. Conventional techniques (i.e., polymerase chain reaction (PCR)-based and gel-based methods) for studying DNA samples require prior knowledge of the sequence, either for PCR primers or for attachment to a target DNA. By focusing on known genes, information from other unknown genes is lost. Multiple genes or multiple infectious microorganisms are known to be involved in many human diseases; however, only a fraction of these genes has been identified, even after the sequencing of the human genome and several microorganisms. The functions of many proteins encoded by these genes are unknown . In addition, studies have estimated that there are approximately 4 × 103 to 104 microbial species per gram of soil, but only less than 1% of microorganisms in nature are observable with the standard culturing techniques . These in turn generated a renewed demand for innovative approaches that can quickly, exhaustively, and intelligently detect and classify gene expression profiles. It is possible to extract genomic samples, such as DNA and RNA, from any biological sample, such as soil, water, and biological specimens, without knowing the genomic identities of the samples. The complementary DNA (cDNA) synthesis from messenger RNA (mRNA) is also well established, either using oligo-dT or random primers. Recent advances in the next-generation sequencing (NGS) technology opened the possibility of de novo sequencing without any pre-existing genomic references . NGS can generate very large volumes of short sequencing reads of genomic DNA (gDNA) at markedly reduced prices and faster rates, and the massive data can be reassembled de novo. However, substantial challenges exist for the de novo applications of NGS, including high error rates, massive information technology systems for data processing and storage, etc.; hence, NGS with de novo assembly is still limited to species-specific applications, including bacterial genomes and mammalian bacterial artificial chromosomes [3–8]. Thus, a challenge is to discover more efficient and effective ways to tap the large amounts of genetic information from a biological sample (i.e., gDNA) or all expressed genes in the biological sample (i.e., cDNA from mRNA), and to use that information to detect changes in genomic patterns and classify them.
Design of Biological Memory with in vitro learning and associative recall
In the Biological Memory (Figure 1a), the initial sequences (termed memory tags) are a set of tag oligonucleotide sequences, to which random sequences are appended during their synthesis. The appended random sequences theoretically contain every possible sequence of a given length. The tag oligonucleotide sequences are designed to be independent of each other in that they do not hybridize to each other (i.e., non-crosshybridizing [NCH]) [10–14] (termed NCH tag), and can be used for printing on a microarray slide for output or separating products (e.g., magnetic bead separation) with proper chemical modifications (i.e., amine, carboxyl, or biotin modifications). Also, the NCH tag could be utilized for other types of product separation or output techniques for downstream processing, such as DNA affinity column chromatography or DNA microarray with sequences complementary to the tag sequence attached to the column matrix or microarray slide. With simple and common recombinant DNA operations, such as polymerization, nuclease digestion, and DNA separation (e.g., magnetic bead separation or column chromatography), the system learns the DNA sequences to which it is exposed. These learned sequences can then be stored as a DNA memory. Subsequently, the learned memory stands can be used to “recall” the input sequences, or sequences that are close under hybridization affinity.
Results and discussion
Principle of the Biological Memory
An underlying hypothesis of the DNA-computing-inspired Biological Memory is that the products, which are learned in vitro, represent the entire DNA population in the input genomic sample (i.e., gDNA or cDNA from a biological sample), and that the differences between input sets could be discriminated by separating the output hybridization patterns between the learned products. When it is assumed that there are two input populations (e.g., set A and set B) (Figure 1b), through the in vitro learning protocol, the elements of each input population are stored in the learned products as content-addressable memory structures. Ideally, the learned products from the set A (LPA) and those from the set B (LPB) should contain every elements of set A and set B, respectively. The sets A and B can then be discriminated through DNA hybridization with their learned products using an appropriate detection system, such as microarray, as a part of the associative recall protocol. At a minimum, the Biological Memory should be able to distinguish between snapshots of a biological sample, from the environment or a living being.
One of the unique advantages of the Biological Memory stems from the utilization of random sequences in the initial memory tags. For example, the population of 20-base random sequences (R20) contains 420 (~1.1 × 1012) different 20-base DNA sequences, which well exceed the number of 20-base segments in any genomes known so far (e.g., ~3 × 109 of human genome). Hence, it is postulated that, using the R20, the entire information of any genome could be captured and stored in a DNA-based memory, if the protocol is properly implemented. As the complexity of genomic samples increases, which would likely happen in real-world samples, for example from the environment or a living being, the Memory protocol could be easily adapted by using longer random sequences (i.e., >R20) to capture more information in the complex samples. The storage procedure is called “learning” because the memory DNA acquires information from examples (i.e., the input DNA), and does so without external knowledge of their genomic sequences. Also, incorporating longer random sequences, such as R20 or > R20, each memory tags can obtain high specificity even at ambient temperature as compared to the previously reported random primer method  with random oligonucleotides of 7 to 10 bases.
Furthermore, in the Biological Memory protocol, the genomic information is processed in one massively parallel step, just as searches have been done using DNA for solutions to hard computational problems . Likewise, matching of stored patterns with new input is done in parallel, in one step. Similarity is implemented in vitro by degree of annealing between new input DNA and the learned memory sequences, thus providing a technique for recognizing patterns in different samples and detecting changes without requiring sequencing. By contrast, the established molecular biology techniques would acquire the genomic information through biological samples with known sequences, and then, depend on a conventional computer for processing the data. Even the new de novo NGS requires massive computer-based information processing for data assembly and their storage [3–8]. The Memory has in vitro capability to efficiently implement pattern recognition and interpretation of large amounts of genomic data that are difficult problems for conventional computers. The Memory is not a laboratory technique only to gather data for conventional computational analyses, but uses the massive scale of storage and parallelism of DNA as the computational tool to draw inferences on the entire in vitro knowledge base quickly and efficiently without any knowledge of sequences. This implies that the Memory can reason and extract knowledge in situations that involve both new and unknown information, which is hard to achieve with conventional laboratory techniques and computational analyses. Also, because there is no DNA sequencing, it is more effective and efficient by alleviating the inherent sources of errors, as well as costs associated with sequencing techniques. Moreover, DNA’s large storage capacity is used to store genomic information from a population or community in the sample for subsequent matching. The information is stored in a compact form, and can serve as a database of the status of the biological sample at a given moment in time. In other words, the Memory is capable of capturing global information on all organisms or whole genome gene expressions in biological samples under certain conditions, and recognizing patterns of contrast and commonality at the population or community level. When the learned memory sequences are attached to a DNA microarray, read out (i.e., “recall”) is easily achieved and interpreted as either a positive or negative match. Also, the microarray’s high capacity for multiplexing is an added advantage, which other conventional gel- or PCR-based approaches cannot afford. Thus, a microarray would provide an ideal vehicle to implement the Biological Memory considering the enormous capacity of nucleic acids on each slide/chip. Finally, it should be noted that the microarray-based Biological Memory would also have significant utility as a simple, fast, flexible, and high-throughput non-gel/PCR based technical platform for genome studies, thus reducing the current reliance on conventional PCR and gel-based methodologies.
Validation and characterization of the Biological Memory
An important property to characterize the Biological Memory is the ability of the learning and recall protocols to learn and differentiate different sets of DNA. In this study, the goal was to evaluate and verify the capabilities of the DNA-based Memory protocol, which include validating and optimizing learning of input DNA sequences using R20 and testing recall of the learned sequences, as well as its sensitivity using gDNA from two model bacterial strains, i.e., E. coli K12 and B. subtilis.
Development of in vitro learning protocol and its optimization
The learning protocol includes three key reaction steps (Figure 2a): the memory tag-input annealing (A), the strand extension by Klenow fragment (E), and Exo I digestion of unbound memory tags and inputs (D). After each step, the products are purified by the magnetic separation. When the orders of the reaction steps were compared, the results (Figure 3b) indicated that the order should be A → E → D. When digestion preceded extension, little or no extension occurred (Figure 3b, lane 4). However, after extension (Figure 3b, lane 7), followed by digestion (Figure 3b, lane 10), the amount of extended strands increased substantially and there existed much less unused memory tags compared to the initial tag concentration (Figure 3b, lane 3 vs lane 7). After digestion, most of the excess unbound memory tags were removed and only extended strands remained, whose molecular weights were higher than the memory tag (Figure 3b, lane 10). The length of the extended strands varied as indicated by the smear of DNA bands, but they were well within the range of the input gDNA length (Figure 3a, lane 2 vs. Figure 3b, lane 10). These results imply the successful learning of the input DNA sequences. For the rest of this study, all learning was performed on the basis of the A → E → D reaction sequence.
To further confirm the learning protocol, it was implemented with the following two negative controls, and the results were compared with the learned products with the positive control (i.e., both input gDNA and memory tag): (1) memory tag only without input gDNA (Figure 3b, lanes 6 and 9) and (2) input gDNA only without memory tag (Figure 3b, lanes 5 and 8). The generalized annealing reactions in the learning protocol can be represented as: Input + Memory tag → Input-Input + Input-Memory tag + Memory tag-Memory tag. Among these, only the extended strands from Input-Memory tag complexes are the actual learned products, i.e. LPI-M. The other two, i.e., learned products from the Input-Input (LPI-I) and Memory tag-Memory-tag (LPM-M), are by-products, which should be eliminated. The LPI-I could be easily separated from the desired learned products (i.e., LPI-M) due to the absence of a memory tag component in LPI-I, i.e., no 5′-end modifications such as 5′ biotin. As expected, no products were observed after extension and magnetic separation (Figure 3b, lanes 5 and 8). Thus, the LPI-I after annealing and extension should not affect the yield estimation of the final learned products, even though it could reduce the amount of available inputs for the memory tag annealing during the learning protocol and potentially reduce the final yield of the learned products. However, LPM-M from the memory tag dimer could negatively affect the expected final outcomes, since it showed similar band pattern to the LPI-M but does not have any information from the input. To investigate the possibility of the memory tag-memory tag duplex formation and their extension, the melting temperature distribution of various lengths of random oligonucleotides were calculated using OligoAnalyzer 3.1 under the reaction condition (1.6 μM of oligonucleotides; 10 mM Na+; 10 mM Mg2+; 4 mM dNTPs) (http://www.idtdna.com/analyzer/Applications/OligoAnalyzer/). The minimum estimated melting temperature of the R20 oligonucleotides under the learning reaction condition was around 42.6°C. This means that most of R20 could form a stable duplex at the reaction temperature (i.e., 25°C). The memory tag concentration in the typical reaction volume in this study (50 μL) was 1.6 μM, which is equivalent to 4.8 × 1016 memory tag strands. The number of maximum independent sequences in the R20 is 1.1 × 1012 (i.e., 420) strands. Thus, the number of each unique R20 sequence in the reaction volume is approximately 4.3 × 104. This indicates that the chances of self-hybridizations, especially perfect matches, between the Individual R20 themselves are extremely low (i.e., 1 of 1.1 × 1011). However, there are still chances for partial hybridizations to form incomplete duplexes. According to the melting temperature estimation, around 7 bases could form a duplex under the reaction condition, since the estimated mean melting temperature of 7-base duplex was 25.7°C, and maximum melting temperature of 5-base duplex was 30.9°C, showing the possibility of the memory tag-memory tag annealing and extension through their partial hybridizations. The length of LPM-M is depending on the location where hybridization occurs and estimated to be around 5 – 80 bases (see Additional file 1: Figure S1), as confirmed by the experimental result using only memory tags (Figure 3b, lanes 6 and 9). When comparing LPI-M and LPM-M (Figure 3b, lane 9 vs. 10), LPH in LPI-M not only had similar molecular weights as the inputs but also exceeded the possible length of LPM-M; thus, LPH should be true learned products. However, the length of LPS was around 50 to 80 bases, which overlaps with the estimated molecular weight distributions of LPM-M. Thus, LPS of the learned products might be the mixture of LPI-M and LPM-M. Nonetheless, it was noted that the concentration of LPS in LPI-M (Figure 3b, lane 10) increased substantially (>5 fold according to the gel intensity analysis) as compared to the negative control (Figure 3b, lane 9). This implies that LPM-M might be present in the learned products, but its concentration would not be significant. The probability of memory tag-memory tag hybridization in the presence of input DNA should be lower than that with memory tag only; hence, the concentration of LPM-M in the learned products should be lower than the memory tag only.
Associative recall with microarray: proof of principle
The capabilities of the associative recall procedure were tested using a microarray detection platform with the learned products from the two model bacterial strains, i.e., E. coli and B. subtilis, which, according to our preliminary analysis, were shown to be genomically very different (see Additional file 3: Table S1). For example, less than 6% of 14-base sequences were common in the digested gDNA of both strains. Therefore, the number of common sequences at the length of the learned products (i.e., >50 bases) must be very low between two bacterial strains.
To analyze the signal intensity difference, BSI and signal-to-noise ratio (SNR) values were extracted using GenePix® Pro 6.0 at three different PMTG values as sensitivity settings, i.e., 800, 900 and 1000. However, the microarray image analyses showed that the signal intensity of spots could not be compared and analyzed statistically, because of the low SNR values of features. To obtain a reliable signal intensity value, for example, to estimate a limit of detection (LOD) and a limit of quantification (LOQ), the SNR value should be higher than 3 [17, 18]. However, even though they could be visually discriminated as described above, most of features showed SNR values lower than 3 (see Additional file 4: Table S2). The low SNR, i.e., low signal intensity, of each spot could be related to the probe length and/or the low discrimination power of the microarray . More uniformly digested and shorter DNA would allow increasing the efficiency of the probe-target hybridization on the microarray. The discrimination power by microarray should be also further enhanced. Increasing the number of fluorescent dyes per probe strand would be one option to increase the signal intensity. Exploring other higher resolution detection system for the recall protocol would be another option, such as the Luminex® system, which is a fluorescent bead-based multiplex assay system for the quantitation and detection of biomolecules, including DNA.
Furthermore, to realize the potential of the Memory protocol that was demonstrated in this study, the protocol should be further generalized to validate its ability to distinguish among different species and strains, and their levels and patterns of gene expression. A microarray could be generated that has spots corresponding to the learned memory strands of gDNAs or cDNAs from different species, strains, or combinations of organisms. Each spot represents the learned product of an organism or a collection of organisms. Unknown biological samples (e.g., gDNAs or cDNAs from real-world samples such as soil, water, and biological specimens) are learned and their learned memory products are recalled with the gDNA or cDNA microarray. Based upon the differences in the hybridization patterns of the learned products, inferences could be made as to the genomic contents and states of the unknown biological samples without their explicit genomic information. Moreover, the technical platform of the recall protocol could be further refined to make the Biological Memory more practical by using oligonucleotide arrays with specific oligonucleotides in known locations on the chip. The Affimatrix GeneChip® System (Affimatrix, Inc., Santa Clara, CA) could produce an oligonucleotide chip with ~9 × 105 different oligonucleotide spots, each of which contains millions of copies of a specific oligonucleotide with a specific length, via the company’s proprietary light directed chemical synthesis process. With such oligonucleotide arrays, hybridization signatures of the learned memory products of unknown biological samples (e.g., gDNA or cDNA from the environment or a living being) could be captured and classified. Comparative analyses of the hybridization patterns and signal intensities of the learned memory products at different time intervals or locations could allow us to effectively sense changes in the genomic state of a biological system of interest and make inferences about the biosystems without their clear genomic information, realizing the promise of the Biological Memory. An added benefit of the oligonucleotide arrays would be that we could acquire sequence information on the learned products on the basis of the known sequence information of oligonucleotides on the array, further expanding the utility of the Biological Memory. Hence, with further improvements as the technology progresses, the Biological Memory protocol may catalyze a paradigm shift in biosensing, from the biomonitorings of ecosystems to the diagnoses of diseases, by allowing the assessment of the large amounts of genetic information or all expressed genes in in situ biological samples at the population or community scale.
This study developed and verified an in vitro Biological Memory to capture and store genomic information from biological samples, both known and unknown, and to classify and compare their genomic patterns. By processing genomic information in vitro, rather than in silico, the advantages include massively parallel sampling of the input DNA, ability to work with unknown organisms and sequences, and massively parallel recall and matching of DNA sequence content to detect changes and classify them. Experimental results with two model bacterial strains demonstrated that the protocols worked as designed, and were able to resolve small differences in sequence content. Specifically, the developed learning protocol was simple, fast, and flexible and could effectively capture the genomic contents of samples without their explicit genomic information at the population or community scale. Also, the recall protocol could discriminate genomic patterns with very fine level of resolutions by template matching reactions between the learned memory molecules and new inputs using the microarray detection platform. However, further improvements are required to fully realize its potential. The learning protocol should be further optimized and generalized to increase the learning capacity, and the discrimination power of the recall protocol by microarray should be further improved. Particularly, it should be not only validated using more complex genomic samples, but also tested and optimized for gene expression studies (i.e., using cDNA). With further improvements, optimizations, and generalizations, it is expected that not only the excellent promise of the Biological Memory system could be realized as the genomic pattern classifier, but also its applicability could be expanded to other various biological and biomedical as well as computational applications. The key point is that biological information is processed without explicit knowledge of its sequence content in order to detect relative changes between samples. Examples include an environmental monitoring method that could provide a holistic view of the genomic status of an ecosystem, a screening tool for the prognosis of human diseases, such as cancer and bacterial or viral infectious disease, a biosensing platform for pathogens in water and food sources, and large-scale DNA-based memories for massive storage and retrieval of non-biological information.
The initial memory oligonucleotide with chemically modified 5′ end, 20-base NCH tag sequence, and 20-base random sequence was designed and purchased from IDT DNA technology Inc. (Coralville, IA): 5′-/Bio (or Am)/GAA AAA ACA CCC CTT CGA TGN NNN NNN NNN NNN NNN NNN N-3′ (mole. wt. 12,668.4 g/mole). To ensure its purity and full randomization, it was purchased with the options of high pressure liquid chromatography (HPLC) purification and hand-mixing. The NCH tag sequence was carefully designed to minimize undesirable cross-hybridization and experimentally confirmed by the previously established methods [10, 13].
The gDNA of E. coli K12 and B. subtilis was extracted by the phenol extraction and ethanol precipitation methods described elsewhere . The extracted gDNA was further digested enzymatically to fragments with an average length of around 200 bases to minimize the formation of secondary structures by the long gDNA and maximize the efficiency of the learning protocol in Biological Memory. The gDNA digestion was implemented according to the previously established method  with some modifications using deoxyribonuclease I (DNase I), a random endonuclease, which produces single-strand nicks in the presence of Mg2+, randomly cleaving each strand of double strand DNA. Briefly, the reaction mixture contained gDNA (0.2 μg/μL), and pancreatic DNase I (Amersham Biosciences, Piscataway, NJ) (0.002 U/μL) in 10 mM Tris–HCl (pH 7.5) and 25 mM MgCl2. After incubation at 37°C for 90 sec, the reaction was terminated by adding EDTA (5 mM). The digested gDNA was purified by the phenol extraction and ethanol precipitation. The purified gDNA was evaluated by the denatured Urea polyacrylamide gel analyses (8 M Urea, 4% stacking gel, 12% resolving gel, 1× Tris-Borate-EDTA buffer) at 60°C. Also, the concentration of the digested gDNA was evaluated using DU® 800 Ultraviolet/Visible/Near-Infrared spectrophotometer (Beckman Coulter, Brea, CA) at the wavelength of 260 nm.
In vitro learning protocol
The reaction mixture consists of 0.02 μg/μL of the digested gDNA (i.e., the input DNA), and 1.6 μM of 5′-amine or biotin modified memory tag in TAEMg buffer (10 mM Tris-accetate, 1 mM EDTA, 10 mM Mg-accetate) as a final concentration. The mixture was incubated at 95°C for 5 min. After a brief centrifugation, it was gradually cooled down to a designated temperature (typically 25° or 55°C) for 30 min, followed by 30-min incubation at 37°C after adding Klenow fragment (5 U/μL) and dNTPs mixture (10 mM). E. coli exonuclease I (Exo I) (20 U/μL) was added, followed by additional incubation for 30 min. During the optimization of the in vitro learning protocol, the initial memory tags with biotin modified 5′ end was used and the learned products were purified using Dynabeads® M-270 streptavidin (Invitrogen corporation, Carlsbad, CA) according to the manufacturer’s instruction. The purified learned products were evaluated by the denatured Urea-PAGE analyses (8 M Urea, 4% stacking gel, 20% resolving gel, 1× Tris-Borate-EDTA buffer) at 60°C. The band area intensity of the Urea-PAGE image was assessed on the basis of the gel intensity analysis with ImageJ software (http://imagej.nih.gov) . For the microarray-based recall, the learning protocol was implemented using the memory tags with amine modified 5′ end.
Associative recall protocol with microarray
The learned products with 5′-amine modified memory tags was purified by the phenol extraction and ethanol precipitation, resuspened in 1× microarray printing buffer (50 mM sodium phosphate [pH 8.5]) as a stock solution (1 μg/μL), and serially diluted for printing with the printing buffer. The 30 μL of the learned products, a negative control (i.e., learned products without input DNA), and a blank (i.e., 1× printing buffer only) were transferred into a 384-well plate. Each sample in the well plate was printed with 10 replications onto CodeLink™ activated microarray slide (Amersham Biosciences) using MicroGrid II microarray printing system with BioRobotics MicroSpot 2500 pins (Genomic Solutions, MI) at 40% relative humidity according to the previously established protocol . After printing, the slides were stored in a customized hybridization chamber (GeneTix, CA), which was filled with saturated NaCl solution at bottom, for 15 hr at ~75% of relative humidity to couple the amine group at the 5′ end of the learned products to the activated N-hydroxysuccinimide-ester group on the microarray slide surface. For blocking, the slide was kept in a pre-warmed blocking solution (0.1 M Tris, 50 mM ethanolamine [pH 9.0]) at 50°C for 30 min. The slide was briefly rinsed twice with deionized water and washed with pre-warmed 4× saline-sodium citrate (SSC) buffer with 0.1% sodium dodecyl sulfate (SDS) for 30 min at 50°C with gentle shaking. The slide was dried by centrifugation and stored at ambient temperature. To evaluate the recall protocol, each digested E. coli and B. subtilis gDNA was labeled using the ULYSIS™ Alexa Fluor 532® Nucleic Acid Labeling Kit, according to the manufacturer’s instruction (Invitrogen corporation, Carlsbad, CA). The ULYSIS™ Nucleic Acid Labeling allows a fluorescent dye to react with the N7 of a purine base (i.e., A or G) in a nucleic acid to form a stable coordination complex, implying the high possibility of labeling the entire strands of the digested gDNA. After labeling, the Alexa-labeled E. coli or B. subtilis gDNA was stored in 4× SSC with 0.1% SDS. For the recall, the labeled target DNA was reconstituted in the hybridization solution (7.5× SSC, 37.5% formamide, 0.15% SDS, 0.3 μg/μL bovine serum albumin) as a final concentration of 0.025 μg/μL. The 20 μL of the target DNA solution was applied to the printed microarray slide with LifterSlip™ microarray cover slide. Hybridization was performed at room temperature for 20 hr in the customized hybridization chamber with ~75% relative humidity. After hybridization, it was washed at room temperature twice with 4× SSC containing 0.2% SDS for 5 min each, 1× SSC for 10 min, and twice with 0.2× SSC for 2 min each. Microarray slide was completely dried by centrifugation.
Image acquisition and data analysis
After recall, the microarray slide was scanned with GenePix 4000B (Axon Instruments, Forester City, CA) at 100% of laser power with 3 different PMTG settings from 800, 900, and 1,000 as detection sensitivity settings. Photobleaching was very minimal at the scanner settings. Signal intensities were measured with GenePix® Pro 6.0 microarray image analysis software (Axon Instruments, Forester City, CA) with 10 μm of pixel size as a detection sensitivity. The obtained data were saved as csv (comma delimited) format for data analysis. Background-subtracted intensity (BSI) values were used in the subsequent analyses. Statistical analyses were performed using statistical package R (version 2.8.0) .
This research has been supported in part by National Science Foundation grant numbers CCF 0523858 (J-WK, JC, and RD), CCF 1049719 (RD), CMMI 1235100 and CMMI 0709121 (J-WK and RD), and ECCS 1128660 (J-WK) and Arkansas Biosciences Institute (J-WK and RD). The authors thank Dr. Marty Matlock for his valuable inputs for the research.
- Strachan T, Read A: Human Molecular Genetics. 4th edition. New York: Taylor & Francis, Inc; 2010. 262Google Scholar
- Amann RI, Ludwig W, Schleifer KH: Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiol Rev 1995, 59: 143-169.Google Scholar
- Metzker ML: Sequencing technologies – the next generation. Nat Rev 2010, 11: 31-46. 10.1038/nrg2626View ArticleGoogle Scholar
- Warren RL, Sutton GG, Jones SJM, Holt RA: Assembling millions of short DNA sequences using SSAKE. Bioinformatics 2007, 23: 500-501. 10.1093/bioinformatics/btl629View ArticleGoogle Scholar
- Chaisson MJ, Pevzner PA: Short read fragment assembly of bacterial genomes. Genome Res 2008, 18: 324-330. 10.1101/gr.7088808View ArticleGoogle Scholar
- Hernandez D, François P, Farinelli L, Østerås M, Schrenzel J: De novo bacterial genome sequencing: millions of very short reads assembled on a desktop computer. Genome Res 2008, 18: 802-809. 10.1101/gr.072033.107View ArticleGoogle Scholar
- Butler J, MacCallum I, Kleber M, Shlyakhter IA, Belmonte MK, Lander ES, Nusbaum C, Jaffe DB: ALLPATHS: de novo assembly of whole-genome shotgun microreads. Genome Res 2008, 18: 810-820. 10.1101/gr.7337908View ArticleGoogle Scholar
- Zerbino DR, Birney E: Velvet: algorithms for de novo short read assembly using de Bruijn graphs. Genome Res 2008, 18: 821-829. 10.1101/gr.074492.107View ArticleGoogle Scholar
- Chen J, Deaton R, Wang Y-Z: A DNA-based memory with in vitro learning and associative recall. Nat Comput 2005, 4: 83-101. 10.1007/s11047-004-4002-3MathSciNetView ArticleGoogle Scholar
- Deaton R, Kim J-W, Chen JD: Design and test of non-crosshybridizing oligonucleotide building blocks for DNA computers and nanostructures. Appl Phys Lett 2003, 82: 1305-1307. 10.1063/1.1556557View ArticleGoogle Scholar
- Chen J, Deaton R, Garzon M, Kim J-W, Wood DH, Bi H, Carpenter DP, Wang Y-Z: Characterization of non-crosshybridizing DNA oligonucleotides manufactured in vitro . Nat Comput 2006, 5: 165-181. 10.1007/s11047-005-4460-2MathSciNetView ArticleGoogle Scholar
- Deaton R, Chen J, Kim J-W, Garzon M, Wood DH, et al.: Test Tube Selection of Large Independent Sets of DNA Oligonucleotides. In Nanotechnology: Science and Computation. Edited by: Chen J. Berlin, Germany: Springer; 2006.Google Scholar
- Yu W, Lee JS, Johnson C, Kim J-W, Deaton R: Independent sets of DNA oligonucleotides for nanotechnology applications. IEEE Trans Nanobioscience 2010, 9: 38-43.View ArticleGoogle Scholar
- Kim J-W, Lee JS, Deaton R: Non-crosshybridizing oligonucleotide building blocks for accurate, scalable nanofabrication. In Proceedings of the second IEEE International Conference on Nano/Micro Engineered and Molecular Systems Bangkok, Thailand. New York: IEEE; 2007:784-787.Google Scholar
- Feinberg AP, Vogelstein B: A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal Biochem 1984, 137: 266-267.View ArticleGoogle Scholar
- Lipton RJ: DNA solution of hard computational problems. Science 1995, 268: 542-545. 10.1126/science.7725098View ArticleGoogle Scholar
- Lee JS, Song JJ, Deaton R, Kim J-W: Assessing the detection capacity of microarrays as bio/nanosensing platforms. Biomed Res Int 2013, 2013: 310461.Google Scholar
- Lee JS, Song JJ, Deaton R, Kim J-W: Exploring the potential of microarray for bio/nano sensing. In Proceedings of the fourth IEEE International Conference on Nano/Micro Engineered and Molecular Systems:5–8January 2009; Shenzhen, China. New York: IEEE; 2009:1065-1068.Google Scholar
- Kim J-W, Carpenter DP, Deaton R: Estimating the sequence complexity of a random oligonucleotide population by using in vitro thermal melting and C o t analyses. Nanomedicine 2005, 1: 220-230. 10.1016/j.nano.2005.06.003View ArticleGoogle Scholar
- Schneider CA, Rasband WS, Eliceiri KW: NIH image to ImageJ: 25 years of image analysis. Nat Methods 2012, 9: 671-675. 10.1038/nmeth.2089View ArticleGoogle Scholar
- Development Core Team R: R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2005.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.