by John McCaskill
Biomolecular computing, computations performed by biomolecules, is challenging traditional approaches to computation both theoretically and technologically. Often placed within the wider context of natural or even unconventional computing, the study of natural and artificial molecular computations is adding to our understanding both of biology and computer science well beyond the framework of neuroscience. The papers in this special theme document only a part of an increasing involvement of Europe in this far reaching undertaking. In this introduction, I wish to outline the current scope of the field and assemble some basic arguments that biomolecular computation is of central importance to both computer science and biology. Readers will also find arguments for not dismissing DNA Computing as limited to exhaustive search and for a qualitatively distinctive advantage over all other types of computation including quantum computing.
The idea that molecular systems can perform computations is not new and was indeed more natural in the pre-transistor age. Most computer scientists know of von Neumanns discussions of self-reproducing automata in the late 1940s, some of which were framed in molecular terms. Here the basic issue was that of bootstrapping: can a machine construct a machine more complex than itself?
Important was the idea, appearing less natural in the current age of dichotomy between hardware and software, that the computations of a device can alter the device itself. This vision is natural at the scale of molecular reactions, although it may appear utopic to those running huge chip production facilities. Alan Turing also looked beyond purely symbolic processing to natural bootstrapping mechanisms in his work on self-structuring in molecular and biological systems. Purely chemical computers have been proposed by Ross and Hjelmfelt extending this approach. In biology, the idea of molecular information processing took hold starting from the unraveling of the genetic code and translation machinery and extended to genetic regulation, cellular signaling, protein trafficking, morphogenesis and evolution - all of this independently of the development in the neurosciences. For example, because of the fundamental role of information processing in evolution, and the ability to address these issues on laboratory time scales at the molecular level, I founded the first multi-disciplinary Department of Molecular Information Processing in 1992. In 1994 came Adlemans key experiment demonstrating that the tools of laboratory molecular biology could be used to program computations with DNA in vitro. The huge information storage capacity of DNA and the low energy dissipation of DNA processing lead to an explosion of interest in massively parallel DNA Computing. For serious proponents of the field however, there really never was a question of brute search with DNA solving the problem of an exponential growth in the number of alternative solutions indefinitely. In a new field, one starts with the simplest algorithms and proceeds from there: as a number of contributions and patents have shown, DNA Computing is not limited to simple algorithms or even, as we argue here, to a fixed hardware configuration.
After 1994, universal computation and complexity results for DNA Computing rapidly ensued (recent examples of ongoing projects here are reported in this collection by Rozenberg, and Csuhaj-Varju). The laboratory procedures for manipulating populations of DNA were formalized and new sets of primitive operations proposed: the connection with recombination and so called splicing systems was particularly interesting as it strengthened the view of evolution as a computational process. Essentially, three classes of DNA Computing are now apparent: intramolecular, intermolecular and supramolecular. Cutting across this classification, DNA Computing approaches can be distinguished as either homogeneous (ie well stirred) or spatially structured (including multi-compartment or membrane systems, cellular DNA computing and dataflow like architectures using microstructured flow systems) and as either in vitro (purely chemical) or in vivo (ie inside cellular life forms). Approaches differ in the level of programmability, automation, generality and parallelism (eg SIMD vs MIMD) and whether the emphasis is on achieving new basic operations, new architectures, error tolerance, evolvability or scalability. The Japanese Project lead by Hagiya focuses on intramolecular DNA Computing, constructing programmable state machines in single DNA molecules which operate by means of intramolecular conformational transitions. Intermolecular DNA Computing, of which Adleman's experiment is an example, is still the dominant form, focusing on the hybridization between different DNA molecules as a basic step of computations and this is common to the three projects reported here having an experimental component (McCaskill, Rozenberg and Amos). Beyond Europe, the group of Wisconsin are prominent in exploiting a surface based approach to intermolecular DNA Computing using DNA Chips. Finally, supramolecular DNA Computing, as pioneered by Eric Winfree, harnesses the process of self-assembly of rigid DNA molecules with different sequences to perform computations. The connection with nanomachines and nanosystems is then clear and will become more pervasive in the near future.
In my view, DNA Computation is exciting and should be more substantially funded in Europe for the following reasons:
The difference with Quantum Computing is dramatic. Quantum Computing involves high physical technology for the isolation of mixed quantum states necessary to implement (if this is scalable) efficient computations solving combinatorially complex problems such as factorization. DNA Computing operates in natural noisy environments, such as a glass of water. It involves an evolvable platform for computation in which the computer construction machinery itself is embedded. Embedded computing is possible without electrical power in microscopic, error prone and real time environments, using mechanisms and technology compatible with our own make up. Because DNA Computing is linked to molecular construction, the computations may eventually also be employed to build three dimensional self-organizing partially electronic or more remotely even quantum computers. Moreover, DNA Computing opens computers to a wealth of applications in intelligent manufacturing systems, complex molecular diagnostics and molecular process control.
The papers in this section primarily deal with Biomolecular Computing. The first contribution outlines the European initiative in coordinating Molecular Computing (EMCC). Three groups present their multidisciplinary projects involving joint theoretical and experimental work. Two papers are devoted to extending the range of formal models of computation. The collection concludes with a small sampler from the more established approach to biologically inspired computation using neural network models. It is interesting that one of these contributions addresses the application of neural modelling to symbolic information processing. However, the extent to which informational biomolecules play a specific role in long term memory and the structuring of the brain, uniting neural and molecular computation, still awaits clarification.
John McCaskill - GMD
Tel: +49 2241 14 1526