Biotech as the Fastest Pathway to an Assembler

abstract

The potential advantages of nanoscale machinery are widely recognized, but practical plans for developing it from today's technology are rare. The main obstacle to developing it quickly is not a lack of suitable molecular components or methods for connecting them, but simply the "chicken-and-egg" problem of assembling small numbers of molecular building blocks into a specifically designed arrangement, before they are joined together.

The use of materials and methods from biotechnology, and in particular the use of DNA hybridization to guide self-assembly, has great advantages for bootstrapping this assembly problem most efficiently, and has many other advantages for the speedy development of the first nanoscale machinery. This pathway is therefore the fastest route to general purpose molecular manufacturing, including to "assemblers", i.e. molecular robots which can construct products more complex than themselves under program control. Since assemblers will speed up all nanotechnology development, including the extension of molecular manufacturing to new materials and operating environments, this is also the fastest pathway to the full benefits of molecular manufacturing using any materials.

Why is it hard to build nanoscale machinery now?

A general ability to manufacture products at the molecular scale would have enormous benefits, which have been recognized for some time. Furthermore, once nanoscale construction can be performed by programmable robots, the time to try out new designs will be small and improvement will be very rapid. This means that it is worth solving the problem of building nanoscale machinery by whatever means would be fastest, even if this means using a limited set of expensive starting materials, since generalizing the materials and construction methods will be much easier once the bottleneck of having the first molecular machinery is overcome.

Given that motivation and the lack of any molecular robot so far, it is reasonable to ask why the problem is difficult. If "building with molecules" is possible (as it clearly is for biological organisms), why is it hard to develop quickly from current technology? (The benefits were recognized as long ago as 1959 by Feynman, and popularized by Drexler almost two decades ago; the fact that biological organisms make heavy use of molecular machinery has also been widely recognized for many years.)

The reason molecular machinery is hard to build is not a lack of usable molecular components. Machines can be built using components of only a few kinds, which form a structural framework, provide for flexible joints and springs, permit motion to be powered and controlled from outside the machine, and include functional parts such as grippers and other tools specific to the machine's application. For some applications, components with electronic, optical, or catalytic activity are also needed.

For all of these functions, molecules and/or nanoparticles can already be designed and synthesized which can perform them and whose properties are understood well enough for use in designed machinery. For example, the strength and spring constants of DNA and several proteins have been measured, as have the forces generated by some biological molecular motors, by DNA hybridization, and by DNA-processing enzymes. In most cases these properties agree well with theoretical predictions.

Another requirement is that the components can be permanently joined, but this too is not where the problem lies -- it is well known how to form strong chemical bonds between the components mentioned above, including between specific amino acids on proteins and specific molecules incorporated into DNA. Although some reactions proceed slowly or require different solution environments, those that would already be practical to use are sufficient for building primitive machinery.

The chicken-and-egg problem: assembling a few parts in a designed arrangement

The major difficulty in making use of the molecular components and joining techniques we already have is that there is not yet a general way to arrange nanoscale parts into the pattern specified by a machine designer, so that they are joined into the designed arrangement, rather than into a random one or into the one their natural affinities would produce (e.g. by self-assembly). Specifically, we not only lack molecular machinery, we don't even have "hands" which can pick up two molecules and hold them next to one another in a desired relative orientation (in a general way), nor do we have any other way to perform that basic operation.

In spite of this, much of current "nanotechnology research" focusses on incremental improvements to what we can already do, rather than on filling this basic gap.

This problem has the character of a chicken-and-egg problem, since most ways one might imagine for "building the first molecular tools" require first solving the same problem of assembling and joining at least a few molecular parts into a desired arrangement. For example, existing scanning probe microscopes can move a molecular tip in 3 dimensions with sub-atomic resolution. But the tip structure, which at the atomic scale is either the end of a carbon nanotube or the rounded corner of a crystal (whose precise atomic structure is neither known nor controllable), is not able to reliably serve as a reversible gripper. To make it serve that function well enough for use in construction probably requires attaching to the tip either a designed object made of at least a few parts in the proper arrangement, or an individual molecule with more complex designed features than what has been achieved so far.

Which approach can solve the assembly problem most quickly?

There have been a variety of proposed approaches for solving this fundamental problem of assembling molecular-scale parts into a designed arrangement. (Note: the lists of associated individuals and companies shown below are representative, not complete.)

• self-assembly
  • using engineered complementary surfaces of proteins, or other organic molecules [Drexler]
  • guided by DNA hybridization [Seeman; Molecubotics]
• direct manipulation using scanning probes
  • of atoms covalently bonded to a solid surface [Merkle, Michelsen, others]
  • of synthetic organic molecular building blocks [Krummenacker, Zyvex, others]
  • of carbon nanotubes [Ruoff, Zyvex]
• direct manipulation using complex micron-scale tools [Feynman, Zyvex]

There are a variety of other proposals for building general nanoscale structures, including nanoscale lithography, molecular imprinting, scanning probes used as "dip pens" or for electrochemistry, self-assembly of other kinds than those listed above, and others; but none of these solve the basic problem of assembling preexisting molecular building blocks, nor would they build atomically-precise structures, so they don't directly address the most basic problem needed for building machinery.

Specifically, self-assembly is not a sufficient solution unless the properties of molecular building blocks which control their affinity for each other can be designed independently of whatever function the building blocks should provide in the final structure. Thus, self-assembly guided by simple chemical affinity will be able to produce novel materials (some of which will be valuable), but not nanoscales machines. Even protein engineering will have problems in that regard, since redesign of the complementary surfaces of a protein is likely to alter other properties as well; in any case, protein engineering is not yet well enough developed to be used for this purpose, and its progress seems steady but difficult and slow. Similar comments apply to engineering of non-protein organic building blocks with complementary surfaces, except for the special case of DNA.

For these reasons and others, we have focussed on self-assembly guided by DNA hybridization as the most likely means of quickly solving this problem. DNA sequences attached to a molecular building block can be redesigned completely independently of the other properties of that building block, designing complementary sequences is easy, and short DNA sequences provide a large number of sequences to choose from. Our documents DNA Guided Assembly of Proteins and How to create Molecular Building Blocks, describe our early proposals on this method in more detail. More recently we have been working with Prof. Erik Winfree of Cal Tech on a related approach using DNA-based molecular building blocks invented by Prof. Nadrian Seeman of New York University.

(We have thought of different practical approaches as well, but they all use materials and methods compatible with biotechnology, since its other advantages, described below, leave no doubt that it is the best general approach to the rapid development of the first molecular machinery.)

Other advantages of biotechnology materials and methods

There are several other reasons why biotechnology methods will permit the fastest development of molecular machinery:

1. Infrastructure: biotech comes with a huge installed base of usable materials, methods, labs, and skilled personnel, and its lab techniques can be performed relatively quickly and cheaply. This is important since turnaround time on experiments determines the rate at which design iterations can be tried.

2. Simple solutions to certain design problems: several of the problems that the first assembler must solve, such as external provision of power, control signals, and the building blocks for its products, have simple solutions using biomolecules which have no analogs in other systems. For example, DNA sequences can be used for both power and selective control of mechanical actuators, permitting nanomachine designs which have potentially thousands of distinct actuators consisting only of simple arrangements of DNA, any subset of which can be controlled independently from the others. Bernie Yurke of Lucent has demonstrated an actuator compatible with this idea, and we have our own design for a DNA actuator which would generate more useful force (which Yurke has seen and thinks would work). Another example is the use of brownian motion in water for delivering building blocks to specific receptors in a nanomachine.

3. Reduced parts count: since the building blocks can themselves consist of composites of biomolecules produced externally, the assembler itself, as well as the other products it should produce, can have a much smaller parts count than if it was built directly from simpler parts. This simplifies the design and building of the first assembler, and also permits it to reproduce itself even if its assembly operations have a higher error rate and lower speed than would otherwise be permitted.

   (Trying to make the first assembler by precisely positioning atoms or small molecules using a scanning probe would probably require at least 10^5 operations, either with no errors, or with a perfect method of correcting errors; current techniques are far from either ability.)

4. Massively parallel manufacturing: biotech-based molecular construction methods produce huge numbers of molecular assemblies in parallel, in constrast to methods based on manipulation of individual molecular building blocks by scanning probes, which (once developed) could produce only one structure at a time. This massively parallel production makes debugging and analysis of intermediate products or subassemblies much easier.

5. Bootstrapping of assembly: probably the most important reason is the one described in the previous section -- the ability to make the first assembler entirely by the "self-assembly" of its components guided by the specificity of DNA hybridization, thus solving the chicken-and-egg problem of having the first viable way to arrange several nanoscale parts into a specified pattern.

The disadvantages of biotech are unimportant

In comparison, the disadvantages of making the first assembler out of soft materials in water are unimportant:

1. The imprecision of position of every component (due mainly to thermal motion of components joined using soft materials) is not a serious problem, since biomolecule-based building blocks can easily be designed so that the assembler need only bring them into proximity with the right unique partners, with the final joining step aided by self-assembly or by the chemical specificity of the joining reaction.

2. The limitation of the first assembler's early products to those which can be assembled in water can be much more easily overcome by building improved assemblers, after the first one exists, than before molecular machinery is available. For example, given the ultimate goal of making diamondoid structures, one pathway to try is to evolve artificial enzymes which let the initial assemblers make highly-crosslinked organic structures, use those to make machines with rigid parts and a sealable piston, then (still operating in water) open the piston to create a vacuum inside, and do operations inside it using mechanical coupling between the inside and outside of the piston. Another pathway would start by making nanoscale glass structures (intended to become parts of a robot that can operate in vacuum) using the same methods used by diatoms, namely placing water-soluble silicate precursors into shaped compartments maintained at the proper pH level. Such things are not hard to try if one can perform basic machine-building operations at the nanometer scale, even if this is limited to a water-based environment, but are difficult or impossible to try before there is any way to build simple nanoscale machinery.

3. If it becomes important to develop profitable products before the first assembler is achieved, simple nanoscale assembly technologies restricted to water still have potentially profitable applications.