This chapter discusses biomolecular electronics and hybrid devices, as well as the relatively new fields in biocomputing. The Army has become increasingly dependent on computers and electronics to achieve high levels of situational awareness, to implement command and control networks, and to support combat systems on the battlefield. In the future, many computing and electronic devices will consist of biologically derived or inspired materials that will increase their usefulness for Army applications.
One problem in discussing research in these areas is the lack of common nomenclature. Molecular electronics is an interdisciplinary field at the interface between chemistry, electrical engineering, optical engineering, and nanoscience. Molecular electronics is defined as the encoding, manipulation, and retrieval of information at a molecular or macromolecular level. These functions are currently performed via lithographic manipulation of bulk materials to generate integrated circuits. Molecular electronics (which includes both biological and nonbiological molecules) greatly miniaturizes computer circuitry and provides promising new methodologies for high-speed signal processing and communication, volumetric data storage, novel associative and neural networks, and linear and nonlinear devices and memories.
Biomolecular electronics (also called bioelectronics 1 ), a subfield of molecular electronics, involves the investigation of native, as well as modified, biological molecules (e.g., chromophores, proteins, DNA), rather than organic molecules synthesized in the laboratory. Because natural selection processes have solved problems similar to those that must be solved in harnessing organic compounds, and because self-assembly and genetic engineering provide sophisticated control and manipulation of large molecules, biomolecular electronics is a very promising field.
From 1975 to 1995, scientists in the former Soviet Union participated in a government-sponsored program to leapfrog the West in computer technology by exploring protein-based bioelectronics. Many of the anticipated applications were military and may therefore be important to the U.S. Army, but details remain classified. One of the best-known accomplishments of the Soviet project was the development of biochrome, a real-time photochromic and holographic film based on chemically modified polymer films containing bacteriorhodopsin (Vsevolodov and Poltoratskii, 1985; Bunkin et al., 1981). The published photochromic and holographic properties of bacteriorhodopsin stimulated the international research that continues today. The protein bacteriorhodopsin is representative of the potential that proteins may have for future Army applications.
Much of the research in biomolecular protein-based devices has focused on bacteriorhodopsin ( Figure 4-1), a protein discovered in the early 1970s that has unique photophysical properties, as well as thermal and photochemical stability. Natural selection has optimized bacteriorhodopsin for light-to-energy conversion, and the evolutionary process has thus generated a native material that is particularly suited for a number of computer and data-storage applications. Bacteriorhodopsin, which is isolated from a salt marsh archaebacteria called Halobacterium salinarium (also called Halobacterium halobium) that has existed for about 3.5 billion years, maintains its structure and function in temperatures as high as 140°C, a temperature at which most proteins can no longer function (Shen et al., 1993). With genetic engineering, bacteriorhodopsin can be optimized for specific applications.
Bacteriorhodopsin should not be confused with rhodopsin, the protein in the back of the eye that converts light into
1 In a more limited context, the term bioelectronics has been used for electronics intended for medical applications.
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 25
Page 25 4 Electronics and Computing This chapter discusses biomolecular electronics and hybrid devices, as well as the relatively new fields in biocomputing. The Army has become increasingly dependent on computers and electronics to achieve high levels of situational awareness, to implement command and control networks, and to support combat systems on the battlefield. In the future, many computing and electronic devices will consist of biologically derived or inspired materials that will increase their usefulness for Army applications. One problem in discussing research in these areas is the lack of common nomenclature. Molecular electronics is an interdisciplinary field at the interface between chemistry, electrical engineering, optical engineering, and nanoscience. Molecular electronics is defined as the encoding, manipulation, and retrieval of information at a molecular or macromolecular level. These functions are currently performed via lithographic manipulation of bulk materials to generate integrated circuits. Molecular electronics (which includes both biological and nonbiological molecules) greatly miniaturizes computer circuitry and provides promising new methodologies for high-speed signal processing and communication, volumetric data storage, novel associative and neural networks, and linear and nonlinear devices and memories. Biomolecular electronics (also called bioelectronics 1 ), a subfield of molecular electronics, involves the investigation of native, as well as modified, biological molecules (e.g., chromophores, proteins, DNA), rather than organic molecules synthesized in the laboratory. Because natural selection processes have solved problems similar to those that must be solved in harnessing organic compounds, and because self-assembly and genetic engineering provide sophisticated control and manipulation of large molecules, biomolecular electronics is a very promising field. PROTEIN-BASED ELECTRONIC DEVICES From 1975 to 1995, scientists in the former Soviet Union participated in a government-sponsored program to leapfrog the West in computer technology by exploring protein-based bioelectronics. Many of the anticipated applications were military and may therefore be important to the U.S. Army, but details remain classified. One of the best-known accomplishments of the Soviet project was the development of biochrome, a real-time photochromic and holographic film based on chemically modified polymer films containing bacteriorhodopsin (Vsevolodov and Poltoratskii, 1985; Bunkin et al., 1981). The published photochromic and holographic properties of bacteriorhodopsin stimulated the international research that continues today. The protein bacteriorhodopsin is representative of the potential that proteins may have for future Army applications. Bacteriorhodopsin Much of the research in biomolecular protein-based devices has focused on bacteriorhodopsin ( Figure 4-1), a protein discovered in the early 1970s that has unique photophysical properties, as well as thermal and photochemical stability. Natural selection has optimized bacteriorhodopsin for light-to-energy conversion, and the evolutionary process has thus generated a native material that is particularly suited for a number of computer and data-storage applications. Bacteriorhodopsin, which is isolated from a salt marsh archaebacteria called Halobacterium salinarium (also called Halobacterium halobium) that has existed for about 3.5 billion years, maintains its structure and function in temperatures as high as 140°C, a temperature at which most proteins can no longer function (Shen et al., 1993). With genetic engineering, bacteriorhodopsin can be optimized for specific applications. Bacteriorhodopsin should not be confused with rhodopsin, the protein in the back of the eye that converts light into 1 In a more limited context, the term bioelectronics has been used for electronics intended for medical applications.
OCR for page 25
Page 26 ~ enlarge ~ FIGURE 4-1 Simplified protein structures. 4-1a Structure and key intermediates in primary and branched photocycles. 4-1b Structure and key intermediates of bacteriorhodopsin. Note: Maximum wavelengths in parentheses are in nanometers (nm). Lifetimes and temperatures apply to the wild-type proteins only and are approximate. Source: Reprinted with permission from Birge et al., 1999. Copyright 1999, American Chemical Society. a nerve impulse. Rhodopsin and bacteriorhodopsin have some structural similarities but few functional similarities. Rhodopsin is not suitable for device applications because it self-destructs after absorbing light. In contrast, bacteriorhodopsin can convert photons into energy, undergoing structural changes once every few milliseconds, and it can do this hundreds of millions of times before it becomes denatured. Russian scientists first brought the device applications for bacteriorhodopsin into focus (Bunkin et al., 1981; Vsevolodov and Poltoratskii, 1985). Recently, other photosynthetic proteins have also shown great potential (Boxer et al., 1992; Lee et al., 1997). The principal Soviet investigator, Nikolai Vsevolodov, has since moved to the United States and is now the principal scientist of Starzent, a small start-up company that hopes to manufacture high-density holographic memories. Vsevolodov’s recent book, Biomolecular Electronics, provides an excellent introduction to the field of protein-based devices (Vsevolodov, 1998). Scientists using bacteriorhodopsin for bioelectronic devices exploit the fact that the protein cycles through a series of spectrally distinct intermediates upon absorption of light. A light-absorbing group (called chromophores) embedded in the protein matrix converts light energy into a complex series of molecular events that store energy. This complex series of thermal reactions causes dramatic changes in the optical and electronic properties of the protein. The excellent holographic properties of bacteriorhodopsin derive from the large change in refractive index that occurs following light activation. Furthermore, bacteriorhodopsin converts light into a refractive index change with remarkable efficiency (approximately 65 percent). The protein is 10 times smaller than the wavelength of light, which means that the resolution of the thin film is determined by the diffraction limit of the optical geometry rather than the “graininess” of the film. Also, bacteriorhodopsin can absorb two photons simultaneously far more efficiently than other materials. Because of this capability, bacteriorhodopsin can be used to store information in three dimensions by using two-photon architectures. Finally, bacteriorhodopsin was designed by nature to function in high temperatures and intense light, a necessary requirement for a salt marsh bacterial protein and a significant advantage for photonic device applications. Bacteriorhodopsin can be genetically engineered to do many different tasks. It can be modified by both random and site-directed mutagenesis, which has opened up opportunities to use this protein in biomolecular-electronic devices. Indeed, some applications have used the protein as a template into which new functionality has been programmed via genetic engineering. When viewed from this perspective,
OCR for page 25
Page 27bacteriorhodopsin is no longer a protein functioning as a photonic device but a complex peptide that can be modified to do whatever one might be clever enough to program into it. It then becomes a prototype for numerous protein-based devices. Optical-Holographic and Three-Dimensional Memories One of the most successful applications of bacteriorhodopsin has been in the development of holographic and volumetric three-dimensional (3-D) memories. The holographic memories take advantage of the large change in refractive index that occurs upon formation of the M state (see Figure 4-1). Thin films of bacteriorhodopsin can generate diffraction efficiencies of about 8 percent, which is more than enough for holographic data storage (Birge, 1992). In addition, because mutants have enhanced its holographic properties, bacteriorhodopsin is competitive with photorefractive polymers but far less expensive (Hampp et al., 1994). Finally, both thin and thick films can be prepared using bacteriorhodopsin in whatever concentration is optimal for the optical architecture. This flexibility is the key to commercially competitive holographic data-storage systems. Holographic memories are considered by some to be 3-D memories because the data are stored as a function of x, y, and θ (the angle of incidence of the write beam). Because this angle is limited by the laws of diffraction to a fairly narrow range, holographic data storage cannot take full advantage of the volumetric storage medium. True 3-D memories can be made, however, by using proteins. One example is the branched-photocycle memory developed at the W.M. Keck Center for Molecular Electronics at Syracuse University (Birge et al., 1999). This memory stores from 7 to 10 gigabytes (GB) of data in a small 1 × 1 × 3-cm3 cuvette containing the protein in a polymer matrix ( Figure 4-2). Data are stored by using a sequential pair of one-photon processes, which allows the use of inexpensive diode lasers to store one bit into the long-lived Q state. The read/write/erase process is fairly complicated, and the reader is referred to Birge et al. (1999) for a detailed description. The Army Land Warrior Program is scheduled to provide each combat soldier with a wearable computer to assist with the processing of sensor and targeting data, situational awareness displays, and communications. As the use of graphical formats to facilitate the assimilation of information in real time increases, the Army will have a growing need for computer memory capacity on the battlefield. In principle, an optical 3-D memory can store roughly three ~ enlarge ~ FIGURE 4-2 Schematic diagram of a protein-based, 3-D memory capable of storing between 7 and 10 gigabytes of digital, error-corrected data in a rugged 1 × 1 × 3 cm3 polymer cuvette. These inexpensive cuvettes can withstand virtually any condition that a human can withstand, including submersion in water for extended periods of time. Selected components are labeled as follows: AMLC SLM = active matrix liquid crystal spatial light modulator; CCD = charge-coupled device. Source: Reprinted with permission from Birge et al., 1999. Copyright 1999, American Chemical Society.
OCR for page 25
Page 28 orders of magnitude more information in the same size enclosure than a two-dimensional optical disk. In practice, because of limitations in optic reliability, the improvement is an approximately 300-fold increase in storage capacity, which is still significant. Protein-based memories have an additional advantage in that the memory medium is extremely rugged. The 10-GB data cuvettes used in the 3-D memory described above can withstand substantial gravitational forces and are unaffected by high-intensity electromagnetic radiation and cosmic rays. Another important advantage of bioelectronic memories is low cost. Protein-based polymer cuvettes for 3-D data storage can be manufactured for only a few dollars. They are also lightweight and insensitive to external moisture; they can even be submerged under water for months without compromising the reliability of the data. Protein-based polymer cuvettes would be a suitable memory medium for troops to carry with them into harsh environments, although extreme temperature can pose problem (boiling water will destroy the data). Directed evolution is currently being studied as a way of improving the thermal capabilities of bacteriorhodopsin. 3-D data storage systems will be developed for the commercial sector, but the Army will need to optimize them for the unique conditions that apply to military environments. Thus, much of the cost of development will be borne by commercial developers, but optimization for military uses must be carried out by the Army. For example, 3-D protein memory cuvettes are already being designed and optimized for archival data storage in office environments; many millions of dollars have been spent on this research and development. Current designs for data cubes are rugged but not rugged enough for possible soldier uses in military environments. In addition, although commercial devices will be inherently resistant to high levels of electromagnetic radiation, they will not be optimized for this important characteristic. Genetic engineering has been used to create bacteriorhodopsin mutants with enhanced materials properties. For example, some mutants have enhanced the holographic properties of the protein by producing an M state with an extended lifetime; others improve the branched-photocycle memory by enhancing the yield of the O state (Gergely et al., 1993; Hampp et al., 1992; Miercke et al., 1991; Misra et al., 1997; Zeisel and Hampp, 1992) (see Figure 4-1). The challenge for materials scientists is to predict a priori which amino acid sequence will create or enhance a specific protein property. At present, most genetic engineering for materials applications is done by trial and error, which reflects the complexity of protein structure and function and the lack of satisfactory molecular modeling tools. It is hoped that continued theoretical research will yield computer programs with predictive capabilities comparable to the SPICE software packages that have become a cornerstone for integrated circuit design. Associative Memories and Processors Associative memories take an input data block (or image) and, independently of the central processor, scan the entire memory for the data block that matches the input. In some implementations, the memory finds the closest match if it cannot find a perfect match. The memory then returns the data block in memory that satisfies the matching criteria or returns the address of the data block so contiguous data can be accessed. Some memories simply return a binary bit indicating whether the input data are present or not. Because many scientists believe that the human brain operates in a neural, associative mode, it is possible that only large-capacity, high-speed associative memories will be capable of leading to genuine artificial intelligence. Researchers have implemented the neural computer memory designed by Paek and Psaltis (1987) using thin films of bacteriorhodopsin as the photoactive holographic media (Birge et al., 1997). The optimization of the holographic properties of the protein via genetic engineering is an attractive aspect of using bacteriorhodopsin (Hampp et al., 1994). Large-scale associative memories and associative processors are considered to be critical components in the development of artificial intelligence. The Army also has pressing needs for processing intelligence and sensor data in visual and other formats from multiple sources in real time. For this reason, important Army applications may be to use associative-memory capability for data fusion and high-speed identification of friend or foe. Rapid, confident decisions could be made if associative processors could be programmed to carry out parallel searches for multiple characteristics. Protein-based associative memories could be implemented in rugged, computer-card format and used in the field to assist in complex decision-making processes. An example of a proposed memory system is shown in Figure 4-3. Artificial Retinas Japanese researchers were the first to develop protein-based artificial retinas (Miyasaka et al., 1992). These retinas displayed excellent motion sensitivity because of the inherent differential responsiveness of oriented films containing bacteriorhodopsin. In subsequent work, the protein films were integrated with charge-sensitive semiconductor circuitry to provide for higher resolution (Chen and Birge, 1993), but this approach was not successful until methods were developed to prevent the protein and the semiconductor surfaces from cross contamination (Tan et al., 1996). Although protein-based artificial retinas are not significantly better than semiconductor versions (i.e., image sensors), they have the potential to be manufactured for a small fraction of
OCR for page 25
Page 29 ~ enlarge ~ FIGURE 4-3 Schematic diagram of a Fourier transform holographic (FTH) associative memory with read/write FTH reference planes using thin polymer films of bacteriorhodopsin to provide real-time storage of the holograms. Note that a partial input image can select and regenerate the entire associated image stored on the reference hologram. Although only four reference images are shown, an optical associative memory can store many hundreds or thousands of images simultaneously. This memory can also work on binary data by using redundant binary logic, and a small segment of data can be used to find the page with the closest largest association with the input segment. Selected components are labeled as follows: AMSLM = active matrix spatial light modulator; CCD = charge-coupled device; FL = Fourier lens; FVA = Fresnel variable attenuator; H1 and H2 = protein-based holographic films; PHA = computer-reconfigurable pinhole array; SF = spatial filter; SP = beam stop. Source: Reprinted with permission from Birge et al., 1999. Copyright 1999, American Chemical Society. the cost of semiconductor versions. Nevertheless, they will require much development before they will be competitive. Commercial production of both types of imagers will be dominated by packaging cost. Artificial retinas are capable of providing nearly diffraction-limited performance and so have great potential for Army high-resolution imaging applications. An artificial retina integrated into a highly sensitive motion sensor, could be incorporated into rugged, inexpensive surveillance pods to monitor enemy movements from a distance. Coupled with an associative processor, these retinas could provide a capability for in situ friend-or-foe determinations. Pattern-Recognition Systems Pattern-recognition systems and associative memories have much in common, including that both systems require a holographic material that is sensitive, is read/write capable, and can operate at or near the diffraction limit. Polymer films of bacteriorhodopsin are the only thick films currently available with these characteristics. German scientists have used holographic thin films of bacteriorhodopsin to make pattern-recognition systems with high sensitivity and diffraction-limited performance (Hampp et al., 1994). A commercial device is available in Europe that is capable of reading paper currency at high speeds, identifying the country of origin and face value, and also identifying counterfeit bills that have previously been read and loaded into the optical reference data banks. These systems are possible because of the remarkable holographic properties of D96N, a mutant of bacteriorhodopsin (Hampp et al., 1992). This site-directed mutant has both an improved diffraction efficiency and an intrinsic frame rate that permits diffraction-limited performance at optimal video rates for
OCR for page 25
Page 30high-speed pattern recognition. The Army should consider use of these systems for target recognition and friend-or-foe identifications. Spatial Light Modulators Spatial light modulators are available in many forms, from threshold devices, which are very simple, to complex optical systems, which impose data on a beam of light. Bacteriorhodopsin has long been used as the photoactive element in spatial light modulators; in fact, this is one of the most successful commercial applications of this protein (Birge, 1992; Birge et al., 1990; Bräuchle et al., 1991; Oesterhelt et al., 1991; Vsevolodov, 1998). At present, standard and holographic spatial light modulators can be purchased from companies in Germany, Israel, and the United States. The Army currently uses this technology in nondestructive testing systems for inspecting artillery and tank ammunition. The apparatus makes use of the diffraction-limited performance of bacteriorhodopsin thin films to achieve a real-time measurement resolution of 0.005mm over a working distance of 25cm (Stuart, 2000). Biomolecular Hybrids Hybrid diodes operating on the principle of photosynthesis are described in Chapter 6 in the section on biological photovoltaics. Other protein-based biomolecular devices include bioFETs (field-effect transistors), which may provide unique architectural opportunities in telecommunications applications, and devices using the photoreactive properties of DNA. BioFETs may enable higher speed operation than conventional transistors, and spintronic injection of semiconductor lattices could produce coherent carrier dynamics, which could be useful for broad-band, fiber-optic converters, and multiplexers. Advances in photonics for telecommunications and radar-signal processing will depend on being able to modulate the phase and intensity of an optical signal. This ability may be enhanced using DNA. DNA-Based Optical-Signal Processing Optical-to-optical interactions based on the photorefractive effect and bacteriorhodopsin offer new and potentially more effective alternatives to crystalline materials such as lithium niobate. Solid-state, single-crystal materials are effective but tend to have limited performance capabilities and to be expensive and relatively inflexible in terms of integration with other devices and materials systems. Electrical-to-optical interactions, which rely on the linear electrooptic effect, are becoming more important as means of modulating optical signals at very high speeds and low power. Recent work performed jointly by Dr. Thomas L. Netzel at Georgia State University and Dr. Bruce Eaton at North Carolina State University suggests that charge migration in DNA may provide the basis for a new class of photorefractive and electro-optic material systems (Rawls, 1999). Netzel and associates have suggested the use of self-assembled arrays of DNA duplexes on the surface of highly sensitive optical wave-guide arrays as a possible photorefractive medium. The approach is based on the attachment of photoactive and redox-active chromophores to 2′deoxyuridine and 2′deoxyadenosine nucleosides. The former would be photoreduced, resulting in electron migration; the latter would be photo-oxidized, providing for hole migration. One of the goals of the proposed research would be to evaluate the effective index change resulting from migration using highly sensitive optical wave-guide interferometers. Assuming that the induced index difference is sufficient, this approach might lead to a very fast optical-to-optical modulation mechanism. Commercial incentives are strong for reducing the cost of optical-to-optical interactions in the commercial telecommunications market, and the Army’s strategic communications infrastructure will be a prime beneficiary. The development of DNA-based optical signal processors is an example of the wide applicability of bacteriorhodopsin and other proteins on which new discoveries in bioelectronics will be based. Technical Enablers and Barriers Biomolecular electronics is many years, if not decades, behind computer engineering but is at the forefront of biotechnology. Change, when it occurs, will occur quickly and will result in devices that could be lighter, faster, and possibly cheaper than computer-engineered devices currently used by the Army. The key enablers and barriers to the development of biomolecular hybrid devices are summarized in Box 4-1. The arrows linking enablers with barriers emphasize that an enabler can only be fully implemented if the linked barrier(s) have been addressed. Two technology enablers may hold the key to genetic engineering and hence the successful development of bioelectronic hybrid devices, including future sensing devices for the Army. Designed combinatorial libraries allow the preparation of de novo proteins. By choosing the libraries intelligently, one could generate a small group (fewer than 100) of potential proteins, all of which would have a reasonable probability of yielding enhanced properties. Screening this group would be relatively easy. An alternative would be to use directed evolution, in other words, allow the bacteria to generate random mutations in the protein and provide some method of testing the value of the protein while it is still inside the animal. Both of these methods would enable a large number of potential mutations to be tested in a short period of time, provided the screening could be done efficiently and accurately. Therein lies the problem. Genetic selection for the desired properties is difficult, and it is unlikely that a global solution can span a large range of target systems. Thus,
OCR for page 25
Page 31 BOX 4-1 Technical Enablers and Barriers to the Development of Biomolecular Hybrid Devices ~ enlarge ~ further research in both combinatorial libraries and directed evolution will be essential to the future development of bioelectronics and biosensors. BIOCOMPUTING Biocomputation is a hybrid field that combines computer science and biology (1) to build computational models of real biological systems, using the tools and concepts of information science, so that biological systems can be seen from a different theoretical perspective and/or (2) to use biological systems or processes as metaphor, inspiration, or enabler for the development of new computing technologies and new areas of computer science. For purposes of this report, the term biocomputing does not include using computers for analysis or data management in biology, which is called bioinformatics (data handling) or computational biology (simulations). Biocomputation focuses on the hybrid field of computer science and biology, including the computational properties of cells (e.g., genetic regulatory circuits), DNA computation, DNA self-assembly, cellular and DNA logic gates, computer immune systems for combating computer viruses, artificial life (Alife), artificial neural nets, and genetic and evolutionary algorithms. The latter three areas were among the first fields developed combining information science and biology; in fact, artificial neural nets have been in existence since the 1940s. Biocomputing is an important emerging discipline. Not only is it a rich field for scientific inquiry, but it may also give rise to a wide range of novel engineering technologies with considerable industrial potential. An understanding of how genetic information determines an organism’s structure and function may provide insights into the construction of new materials and structures on a molecular scale; conversely, concepts from computer science may shed new light on how the genetic code evolved, and even on the principles of life itself. For example, certain biocomputing practices may offer a viable alternative to semiconductor-based computing, which might circumvent the problems anticipated by Moore’s law in the next few years. (Moore’s law, named for the founder of Intel, predicts that the density of components possible on a computer chip will approximately double every 18 months.) Biological Models Certain biological organisms, such as cells, can serve as models for sensors or digital logic gates. Computer science concepts are helping molecular biologists analyze and simulate genetic regulatory networks by understanding how cells process information. By viewing biological systems as information-processing units with regulatory logic and circuits, the mechanisms, design principles, and dynamic behavior of regulatory networks in cells can be better understood. A related area is the development of cellular logic gates, whereby a cellular “inverter” can be made based on proteins that either suppress or activate the production of other proteins, the suppression/activation mechanism acting as an on/ off electrical switch. Artificial neural nets, another example, are computer algorithms that mimic the way neurons in the brain process
OCR for page 25
Page 32information via the generation and transmission of electrical signals. They are used in a kind of iterative methodology, whereby real data are used to train the artificial neural net to recognize certain patterns (e.g., Ford Motor Company has explored using neural nets to recognize engine misfires). Evolutionary and genetic algorithms are computer programs based on the concept of simulating evolution via the processes of natural selection, mutation, and reproduction. Artificial life, or Alife, is a new discipline that studies “natural” life by attempting to recreate biological phenomena from scratch using computers and other artificial media. Just as synthetic chemistry enables scientists to create chemicals not found in nature, the goal of Alife research is to create biological phenomena in nonliving media. These are only a few examples of the extensive potential for scientific and technological innovation arising from the confluence of biology and computer science. Biological processes are likely to continue to inspire versatile and useful applications. DNA Computing Even though biologically inspired computing technologies may only prove to be useful for very specialized problems, their potential is still impressive. For example, compared to conventional computers, DNA used as a computing medium may prove to be a billion times more energy-efficient and to have a trillion times more data-storage capacity. (DNA stores information at a density of about 1 bit/ nm3, about a trillion times as efficient as videotape.) DNA computing is also massively parallel. Researchers are currently trying to exploit these properties for several purposes, including solving NP-complete problems (mathematical problems whose answers cannot be checked in computer running time bounded by a polynomial solution), searching large databases, solving problems that require vast amounts of memory, and encrypting data. Originated by Leonard Adleman at the University of Southern California in 1994, DNA computation makes use of the encoding properties of strands of DNA subunits to compute the solution to problems. This innovative method of problem solving may be useful for problems that would be intractable by traditional computing methods. Adleman was the first to use DNA encoding and storage properties and manipulate single strands so that they could link up in ways that represent the solution to a problem (Adleman, 1994). DNA consists of two long chains of alternating phosphate and deoxyribose units twisted into a double helix and joined by hydrogen bonds between two pairs of nucleotides, adenine and thymine (A and T) or cytosine and guanine (C and G). In living organisms, each base pair bonds with its complement—A to T and C to G—in a sequence that determines the organism’s hereditary characteristics. To use DNA in a computation, one must first puzzle out which sequences reacting in which ways accurately replicate the algorithm in question and then custom make the single strands of DNA with the desired sequences, known as an oligonucleotides. Typically, a DNA-based computation is arranged as a series of test tubes filled with water and up to 1,020 strands of DNA. Each test tube is created from earlier ones by one of several operations, such as separating the strands by length, pouring one test tube into another, extracting strands with a given pattern, heating or cooling, or using enzymes to splice the DNA. The series of test tubes form a single-instruction-multiple-data computation performed in parallel on the DNA. In 1994, Adleman first successfully tested the theory of DNA computing on the directed Hamiltonian path, or “traveling salesman,” problem. The challenge is to figure out a single route that would take a “salesman” to multiple cities from a given starting point to a given end point, passing through each city once and only once. Using recombinant DNA laboratory methods, Adleman was able to extract the correct answer to the traveling salesman problem out of the many random paths represented by the DNA. Related problems amenable to DNA computing solutions are optimal shop scheduling and the longest path in a graph (Adleman, 1994; Gifford, 1994). Other candidates for DNA computing are cryptography, problems from computer-aided design (e.g., such as checking out the correctness of circuits or protocols), parallel searches, and factoring. Scientists have investigated using DNA computation for encryption and to self-assemble lattice structures on the nanometer scale, which could ultimately be used for nanomanufacturing processes. Other scientists have devised methods of computing with DNA on substrates. DNA has also been used as a data-storage medium, for which associative search methods are being developed for retrieving the encoded information. DNA computing might be used to integrate a complex array of intelligence data into a strategic plan (Forbes, 2000). DNA computing is not a general-purpose type of computing and is, therefore, inherently limited to the types of problems discussed above. Other weaknesses may also limit the value of DNA computing. For example, individual operations are slow, and set-up time and materials costs associated with complex problems can be considerable. Complex problems may also prove to be volumetrically limited. It takes 0.5 grams of DNA to make 256 strands of 1,000-unit length. To make 270 strands of the same length requires 8 kilograms of DNA. Initial forms of DNA computers are error prone. Nature has long relied on errors in the DNA replication process for evolutionary advancement. Sometimes these errors enhance the organism, but in most cases they damage an organism and lead to its early demise. Errors are sufficiently common that evolution has spawned an entire protein-based process to monitor and correct errors that occur during replication. Current DNA computing does not take this tendency into account, which could lead to erroneous results.
OCR for page 25
Page 33 In summary, early research in DNA computing shows it to be inflexible with respect to application domain and expensive in terms of logistic demands for space and materials. Newer applications, based on such things as programmed mutagenesis or using DNA as scaffolding, may lead to more adaptable DNA computing algorithms. Overall, however, the possible benefits appear to be outweighed by likely problems in field applications. The Army should continue to monitor basic research in this field, however, because new methodologies might yield new paradigms that permit faster, less expensive, or more general applications. KEY RECOMMENDATIONS As the Army becomes increasingly dependent on semiconductor electronics it becomes ever more vulnerable to the effects of radiation and extreme electromagnetic pulses associated with detonations of nuclear or other high-radiation weapons. Unprotected electronic components, which are critical to the Army’s command and control, communications, computers, intelligence, sensors, and reconnaissance (C4ISR) capabilities, are especially vulnerable. Furthermore, there is a limit to how well semiconductor electronics can be protected from electromagnetic radiation because the two key schemes used to protect them (high redundancy and Faraday isolation) add significant weight and increase power consumption. The development of biomolecular hybrid components may reduce this vulnerability to radiation extremes. The extent to which bioelectronic components are inherently insensitive to radiation has not yet been fully explored. Because none of the mechanisms responsible for electromagnetic-induced catastrophic failures of semiconductor devices would be active in biomolecular electronic devices, logic would suggest that these devices would exhibit high tolerance. Clearly, this biotechnology is important to the Army, but has limited nonmilitary applications. Therefore research funding will have to be supplied by the Army. Proteins are the essential components of protein-based materials and devices that will result from advances in molecular electronics. The value of these devices to the Army will increase significantly, as the proteins are improved and optimized for specific bioelectronic and biosensor applications. Protein developments could also be valuable in the development of new threat agents by potential adversaries; thus, they are potentially important to defense against chemical and biological agents. For these reasons the Army should closely monitor commercial and academic developments to optimize protein characteristics through discovery and genetic engineering. Protein-based devices that have already demonstrated some potential will require support from the Army to ensure future developments that meet Army needs. The Army should support the development of protein-based data-storage and associative-memory devices, which have been identified as being well suited to meeting military requirements for rugged data-storage media and data-fusion applications, respectively. Army sponsorship will be necessary to ensure that these devices are optimized for use in the field. The extent to which biomolecular hybrid components are inherently insensitive to radiation extremes has not been fully explored. None of the mechanisms responsible for electromagnetic-induced catastrophic failure within semiconductor devices would be active in biomolecular electronic devices. The vulnerability of C4ISR systems is of critical concern, and the Army should support research to determine the extent to which bioelectronic components are resistant to radiation-induced failure.