Copyright (C) 2000 by Bruce Smith, Markus Krummenacker, and Jim Lewis (Molecubotics)
Outline: 1. Goal: biotech assembler. 2. Requirements: 1. Assembly of MBBs (before an assembler is available) 2. MBBs for various functions 3. Grasping tool 4. Attachment tool 5. Macroscopic interface 6. Design of assembler systems 3. Technical Strategy and Milestones (most can be worked towards in parallel): 1. Assembly Milestone: controlled assembly of small aggregates of MBBs. 2. Scaffolding Milestone: assembly of hundreds of MBBs. 3. Actuator Milestone 4. Simple Machine Milestone 5. Grasping Tool Milestone 6. Attachment Tool Milestone 7. Simulated Assembler Milestone 8. The First Biotech Assembler 4. Potential Spin-off Products 5. Project Organization ---
The goal of Molecubotics is to develop the first assembler, using techniques taken mainly from biotechnology.
Definition and importance of an "assembler":
The key bottleneck on the path towards realizing the full economic promise of nanotechnology is to build an "assembler": a molecular robot that can be programmed to construct molecular machinery made from the same kinds of parts as it is (which can therefore construct copies of itself).
Having a working assembler system, even if it is very limited in the kinds of building blocks it can use, permits the cost of molecular machinery to drop to the cost of raw materials plus design, and permits rapid development of improved assemblers and of many other new products.
Therefore, to develop the first assembler, choosing the fastest and simplest strategy is much more important than choosing the kind of "building blocks" used in the most desirable eventual applications. Once assemblers are in hand, they can quickly be made to use better building blocks if necessary.
Our analysis of currently available approaches (as summarized in Biotech as the fastest pathway to an assembler) concludes that starting with biotech methods is the easiest approach, all things considered.
A biotech-based assembler of the simplest kind is envisioned to consist of roughly 100 to 200 MBBs (Molecular Building Blocks), using an "alphabet" of 10 to 15 different types of MBBs. Some MBB types would be incorporated in many places in the assembler structure. Having more MBB shapes available can permit assembler designs containing fewer MBBs, especially after MBBs usable as "rigid rods" are available.
Such an assembler would also need several (2 to 5) different "control channels", i.e. parts whose state or position can be independently modulated by outside signals sensed in parallel by all the assemblers dissolved in one solution or attached to one surface. More complex molecular machinery, developed later, could include control logic permitting a single outside signal to control a population of assemblers, or permitting internal control by on-board programs.
Development of the first biotech assembler requires solving several problems. We describe the problems here, and our strategies for solving them in section 3.
Ability to assemble MBBs (Molecular Building Blocks) into precise, complex, designed arrangements with up to hundreds of distinct component positions, with a quick turnaround time for building newly designed structures from the same kinds of MBBs.
Several kinds of MBBs (synthetic constructs of modified proteins, DNA, and/or small organic molecules), able to be assembled as in 2.1, and able to serve the following functions:
The functions above are sufficient to construct an assembler; many potential earlier products would require only some of these functions. For every required MBB function, there are known biological or organic molecules able to serve it, when slightly modified using known techniques.
Some of the biomolecules able to serve in MBBs with various functionalities are: <table not yet here, but we have it elsewhere ###>
The application-specific MBB functionalities required by an assembler (grasping and MBB-attachment tools) are given separate requirements sections below.
Farther ahead, components which support additional functions would be useful:
Any assembler needs a way to hold onto its "workpiece" (the MBBs it has already joined together to form part of its product), and to actively move new MBBs into the desired position and orientation relative to the workpiece, before attaching them to it.
Holding the workpiece is unnecessary if both it and the assembler are bonded rigidly to some substrate. It is more likely that the first assembler will either be fully dissolved, or will be insufficiently rigid for this to work. In that case, the assembler will need to move the new MBB relative to the workpiece while holding both of them.
Depending on the relative strength of the grasping bond, bonds within the workpiece, and the assembler's actuators, the tools for grasping either the new MBBs or the workpiece may need to have a grasping affinity which can be modulated in some manner.
After an assembler moves a new MBB to its desired place in the workpiece, it must attach it to the workpiece before it can place more MBBs. This requires that the MBB stick more strongly to the workpiece than to the grasping tool ("hand") which is holding it, so that it stays in place when the hand is moved away.
If the new MBB-workpiece bond is not strong or robust enough for the final product, it must also later be strengthened or augmented by a better bond, but this might be done as a separate processing step for the product as a whole, rather than by the assembler per se. That is, there is a basic choice of whether MBB attachment is a 1-step process (where the assembler makes the permanent MBB-workpiece bond before releasing each MBB) or a 2-step process (where the assembler only makes a temporary bond, good enough to allow it to add more MBBs to the workpiece).
The choices for MBB attachment bonds and MBB grasping tools (hands) interact strongly due to the need for the hand to leave the MBB on the workpiece when it moves away. This is the issue most likely to favor making an immediate permanent bond between MBB and workpiece, since then (as described in section 3.5) the hand can probably just pull away without down-modulating its affinity for the MBB. However, making permanent bonds quickly is a nontrivial problem, as discussed in section 3.6 (Attachment Tool Milestone). So the first assembler might make use of either of two basic schemes:
(Note that the first scheme is similar to the "DGAP" scheme (section 3.1.1) for using DNA-guided self-assembly to arrange MBBs before an assembler is available, followed by a non-specific slow reaction to strengthen all MBB-MBB bonds at once.)
See section 3.6 (Attachment Tool Milestone) for discussion of how to meet this requirement.
A useful assembler system consists of a large number of discrete molecular machines (assemblers), constructed as in 2.1 from parts in 2.2, mounted or enclosed in some manner which permits:
The main choices for the operating environment of a population of the first kind of assemblers will be:
The latter choice should permit viewing of the assemblers and their products by an AFM operating in solution, and it might even permit influences by the scanned probe on individual assemblers (mainly useful for debugging).
The first assemblers will probably be dissolved in a fluid, or attached at random positions to a common substrate, but later ones may benefit from being attached to each other and organized at larger spatial scales, and from containing different specialized machines working together, some of which perform internal computations which control the others.
This is why we speak not just of "assemblers" but of an "assembler system", which is like an "operating system for molecular machinery" (analogous to a software operating system), and also somewhat like an "artificial cell or organism".
The design of an assembler system will ultimately be much more complex than that of each individual machine it contains. But the simplest systems (identical assemblers controlled externally and in parallel) will work well enough to get through the "assembler bottleneck" (section 1) and greatly speed up further development.
One requirement is to design the first assembler system, including all the assembly instructions needed to make the first assemblers "from scratch", and to design the control signal sequences which cause those to make more assemblers, other products, or more advanced assemblers.
Since even the first assembler may contain a few hundred MBBs per copy, its design is likely to benefit from both off-the-shelf and custom-built software tools, for CAD and for analysis and simulation at various levels, covering not only atomic structure and chemical properties of components in detail, but also mechanical and thermodynamic properties of components treated as units. Development of this software is also a significant requirement, especially since improvements beyond the very first assembler are more likely to be limited by the speed of generating new designs to try than by the speed of instructing the existing assemblers to build them.
As explained in our document Biotech as the fastest pathway to an assembler, finding methods to assemble simple aggregates from well-defined MBBs with high spatial and geometric control is the most fundamental and important problem that is holding back substantial progress in nanotechnology. Thus, this milestone is the most urgent. It will be achieved when:
(1) At least one kind of MBB has been developed, which can be varied in small ways (i.e. very similar MBB types can be developed) in a rapid and reliable manner; the MBB might consist of a simple synthetic construct of a specific protein, some DNA, and perhaps small organic groups, or it might be a pure-DNA construct such as a "double crossover molecule".
(2) MBBs of this kind can be reliably assembled into small, well-defined aggregates, containing 5 to 10 MBBs each, in a precisely controlled arrangement, which can be altered by design in a straightforward and general manner.
Since most development work awaits this basic ability to assemble small structures, several methods should be tried in parallel. Some of the methods we think can be most quickly developed include:
This requires each structural MBB to have different DNA sequences at several specific locations on its surface, as well as chemical groups which can be covalently bonded to lock the arrangement of MBBs into place once it has self-assembled by DNA hybridization. The same MBBs, but with different DNA sequences, would form different structures.
(DGAP is the subject of a 4-page "description of invention" notarized in January 1994, and is described in detail in our 1996 document A Proposed Path from Current Biotechnology to a Replicating Assembler; several possible ways to make MBBs suitable for DGAP are described in our 1997 document Production Strategies For Molecular Building Blocks Suitable For DGAP. One of the simplest ways makes use of biotinylated DNA attached to the protein Streptavidin; we have computed specific DNA sequences which should permit this method to work. Unlike other proposals for joining MBBs using Streptavidin and biotin, or indeed unlike most proposals for using DNA to build nanostructures, we propose a method for selecting only one of the several possible arrangements of 4 different DNA sequences in the 4 locally identical binding sites on Streptavidin; this is crucial for obtaining specific structures when these MBBs are used in complex aggregates.)
Seeman and Winfree have shown that periodic 2-dimensional arrays of DX molecules can be formed by DNA hybridization. In principle, it should be straightforward to extend this procedure to more complex patterns, to bounded rather than periodic patterns, to 3 dimensional arrays, and to use chemically modified DNA with more desirable properties.
(For double crossover (DX) molecules, this milestone is close to having been already achieved by Seeman & Winfree; however, for many uses it will be necessary to modify DX-based MBBs to give them improved strength, altered chemical structure, and/or attachment points for other kinds of MBBs based on proteins. For this reason, even after achieving this milestone with DX molecules, it may be worth pursuing it with other kinds of MBBs.)
There are a variety of other possibilies to explore, which we have not yet studied in as much detail. For example, some kinds of proteins can form 2D or 3D crystals, accessible by scanning probe (AFM), to which other MBBs could be weakly attached, and then removed from selected positions by some modification performed by a scanning probe.
Once small aggregates of MBBs can be routinely put together, this should be scaled up to larger structures consisting of up to hundreds of MBBs. Depending on how well the small aggregate technology works, this may require optimization of MBB purity, of MBB-bonding yields, of assembly sequence or speed, and/or of stability of MBB bonds (e.g. by adding additional crosslinks).
It may also be desirable to develop a wider range of MBB shapes (especially, long rigid rods) to reduce the parts count of structural MBBs required for the scaffolding or framework of each assembled structure.
This milestone will have been achieved once large aggregates can be routinely synthesized and characterized.
This milestone can be worked on in parallel with the development of more kinds of MBBs and more uses for small aggregates. If the large aggregate milestone takes too long to achieve, there are ways to develop an assembler which don't require them. However, the selectivity of DNA hybridization is in theory sufficient to permit thousands of parts to be joined in a unique arrangement.
Any interesting machine needs movable parts and ways of moving them, i.e. actuators and/or motors. Developing and testing at least one actuator usable in an assembler is another important milestone.
There are several actuation methods possible using DNA, some of which have been published by others and one of which is (to our knowledge) proprietary to us. There are several protein-based molecular motors that the research community is studying in ever more detail, all of which could potentially be used in machine designs. There are a variety of proteins and protein domains whose shape is changed by binding of small ligands. Other methods are mentioned below.
A requirement for an actuator to be useful in a primitive assembler, often not appreciated, is that multiple actuators in one device can be controlled with independent external signals. For example, if proteins which change shape in response to small ligands are used, several *different* proteins (which respond to different ligands) will be needed in each machine, to provide it with enough independently controllable degrees of freedom. This suggests that actuator designs which use specific DNA sequences for control will be the simplest to develop for this purpose, since only a single design (with a DNA sequence which can be easily varied) must be developed.
Achievement of the assembly milestones (3.1 and 3.2) will help greatly with building and testing actuators, since they need to be anchored in a structure, much like muscles must be anchored to a skeleton. Some results might be achieved earlier, however, with actuators suspended between scanning probe tips and substrates, or between glass beads suspended by optical tweezers, both of which methods have been used (for example) to measure forces related to DNA hybridization or stretching.
This milestone will have been achieved when at least one actuator sufficient for use to control an assembler (which permits multiple actuators in one machine to be independently controlled) has been constructed and experimentally characterized, and when it is made compatible with the MBB assembly method developed in 3.1 and 3.2. Parameters such as force generation and positioning accuracy need to have been measured at least in a crude manner, along with error estimates for those values. It will also be important to know with what cycle time the actuator can be moved to a different position, and how this might be optimized further if necessary.
More advanced (internally complex) assemblers can take advantage of faster actuators even if fewer independent control signals are possible for them, so control methods other than DNA (which diffuses relatively slowly due to its size) should be explored even if DNA control signals are sufficient for the first assembler. Even the first assembler may benefit from a few fast control channels in addition to a larger number of slower (DNA-based) control channels.
Besides protein shape modulation by small ligands, fast actuation of many machines in parallel might be possible by variations over time of temperature, ionic concentration (including pH), electric field, tension or compression in a substrate, or other variables. (One exception is solution pressure, which doesn't appear easy for biomolecule-based actuators to respond to, since the biomolecules I'm aware of show too little volume dependence on pressure. Eventually it might be possible to use proteins embedded in the surface of a gas-containing vesicle as a pressure-sensitive actuator.)
There are also molecular motors sensitive to variation of ionic concentration across a membrane, and membrane-embedded ion-channel proteins sensitive to the electric field across a membrane. One such ion channel has been engineered (by Micah Siegel) to contain green fluorescent protein whose fluorescence differs when the channel is open or closed; reengineering such proteins in other ways might be used to develop other kinds of actuators or transducers.
There are also possibilities for actuation of individual machines (one at a time) using electric current in a wire built into a substrate, optical tweezers moving micron-sized beads covalently attached by tethers, influences by scanning probe tips, and other methods.
Once an actuator is developed, several can be assembled into one device, which will constitute a simple machine, with a few degrees of freedom of control.
Such a device would probably require 10 to 50 MBBs, and would therefore require prior achievement of the Scaffolding milestone (3.2). This is also the level of complexity at which custom software tools are likely to become especially important for design and simulation.
This milestone will have been achieved when a simple machine can be constructed, controlled, and proven to move as expected when under control.
(Relative positions of parts in each machine can be monitored statistically (averaged over all machines in a large sample being controlled in parallel) by the change in fluorescense of optically active molecules depending on their separation, and perhaps by other methods such as NMR. Direct observation of individual machines by AFM might also be possible, depending on the machine design.)
As described in 2.3, assemblers must grasp both their workpiece and the new MBB they want to attach to it. There are several conceptually different ways this might be organized. The workpiece and the new MBB might be grasped by different methods, using specialized sites on the workpiece, or by the same methods, so the assembler needs only one kind of grasping tool, or "hand" (though it would need two or more instances of that hand in each assembler). New MBBs must be grasped and released many times during a product's assembly, whereas the workpiece might be grasped and released only once -- or the assembler might continually move along the workpiece (especially for workpieces larger than the assembler), or regrasp the workpiece at a new site before adding a few MBBs near that site.
We will focus here on the simplest schemes which might be adequate for the first assembler to make copies of itself, even though the more complex schemes will be very useful for extending what the assembler can build. Accordingly, we'll assume that the workpiece can be grasped on the same kinds of sites and by the same kinds of "hands" as the new MBBs can, so we can limit the discussion to the hands for grasping new MBBs.
One conceptually simple grasping scheme is to have a single hand for grasping new MBBs, whose affinity for a specific part of the MBB can be controlled by an external signal (such as the concentration of a ligand which binds somewhere else on the hand). MBBs added to solution diffuse onto this "hand" and stick there in a controlled orientation (see section 3.6 for how they can avoid prematurely sticking directly to the workpiece); the hand is moved to the desired position using other control signals; the MBB is then joined to one or more other MBBs already in the workpiece (section 3.6); then the hand control signal is modulated to make the hand release the MBB. (Which kind of MBB the hand grabs is determined by adding only one type at a time to the solution.)
The affinity of the hand for the MBB it grasps might be modulated by the concentration of a small ligand which binds to a protein in the hand, by DNA which binds to complementary DNA in the hand, by moving a "fake MBB" built into the assembler into a position where it can compete with the "real MBB" for binding to the hand, or by other means.
A simpler scheme, which avoids any need for modulating the hand-MBB affinity, is for the hand to just pull away from the MBB after it is joined to the workpiece; this requires that the hand-MBB affinity is weaker than both the MBB-workpiece bond and the actuators used to move the hand. We expect that this scheme will be adequate for the first assembler, but even so we will need to compare the extra complexity of affinity-modulated hands with the easier requirements on MBB attachment methods which their use would permit.
Due to the relationship among the choices for grasping method, MBB attachment method (3.6), and the actuators for moving the hand, these components can't be designed in isolation. It is also possible that the first assembler will need two different kinds of hand so it can handle two classes of MBBs which are grasped in different ways. Even so, it is useful to make a separate milestone for developing a grasping tool.
The grasping tool milestone will have been achieved when:
It is worth noting that since the first assembler will necessarily have a structure which can be put together just by self-assembly, it will have an easier job than described here when making a second assembler with the same structure, whose MBBs could after all just self-assemble into the correct places. That is, there is a tradeoff between how much the correct positioning of each MBB in the product is controlled by that MBB (e.g. by the DNA sequences attached to it), and how much needs to be controlled via motion of the assembler's hands.
It may prove simplest to let the first assembler gain much assistance from MBBs capable of self-assembly, since this would permit it to work even with crude positional control of MBBs, and no orientational control; even then, use of an assembler would permit more complex products than could be produced with self-assembly alone. However, as soon as the assembler itself does most of the work of arranging, orienting, and joining the MBBs in the right way, the MBBs themselves can be much simpler, and the objects made from them can be much larger and more complex.
(See requirement 2.4 for the problem to be solved and for two different general strategies for solving it.)
There are many possible methods by which an assembler might attach a new MBB to its workpiece; we don't yet know which one will be simplest to develop, but any of them might turn out to be difficult. Of the problems to be solved, this one has the most uncertainty about the nature of the best solution. Accordingly, it is necessary to study several possible ways:
Problems: if the reaction is fast, MBBs might attach to undesired locations on the workpiece before they became bound to the "hand", but if it's slow, the assembler will need too much time to attach each MBB. There are many reactions to consider, but many of them are likely to be slower than ideal.
Possible solutions (to slowness and/or undesired attachment points):
If DNA is used as a temporary attachment method, it might be possible to use DNA ligase (perhaps tethered to a movable part of the assembler) to make such bonds permanent; note that the effective concentration of an enzyme can be made quite high by holding it in the right place without much positional freedom, which can speed up the reaction rate.
There are chemical groups which can be built into DNA which form permanent bonds when held near each other, as they would be by DNA hybridization; some of these require photoactivation, and others require only time and appropriate conditions (e.g. "Glick bases" which form disulfide bonds). Such groups could be built into an ssDNA strand near the point at which it is attached to an MBB, providing a short connection length between two such MBBs.
Note that several tether-like connections between MBBs can result in a rigid connection if they are properly arranged around a surface on which the MBBs make intimate contact. Also, in many products, many of the connections between MBBs need not be rigid, when all that's required for product operation is sufficient nearness between connected MBBs. This is often the case, even for most components of an assembler (the main exception being an outer rigid framework).
The most attractive connection method in the long run is likely to be the use of catalysts, such as evolved proteins, moved near the reaction site by the assembler, to speed up covalent bond formation requiring no extra reactants. But it's not clear whether this can be developed sooner than one of the less ideal methods.
This milestone will have been achieved when:
If yields are not high enough, products can be designed with redundant connections between MBBs, and/or from subassemblies built in parallel, which can be tested so that only the correct ones are further assembled into final products.
This section is not finished ###. Outline: it is desirable to simulate the operation of assemblers like the one to be built first, and useful even at prior stages, in order to understand the tradeoffs between various MBB limitations, as well as to explain and demonstrate the basic idea of what we are trying to achieve. Since any assembler is too large for atomic-level simulation, it must be simulated mainly at a higher level, more like the level on which its designers or molecular biologists would think about it, corresponding to the mechanical and thermodynamic properties of its components, treated as units.
The real test of the biotech assembler-development methodology will be to build the first machine that can construct copies of itself when controlled by external signals.
This will require a moderately complex design, probably containing on the order of at least 10 actuators and 200 MBBs (though these are very rough estimates, and there are tradeoffs between MBB count, MBB variety, and design software complexity).
The development of the first assembler will use the results of all previous milestones, though most of those should be worked on in parallel, so that design tradeoffs can be made which permit the first assembler to be built sooner.
This milestone will have been achieved when the number of functioning assemblers in our laboratory grows over time, since new ones are constructed faster than old ones deteriorate, even when the yield of harvesting new assemblers and placing them into operation in a new (or the same) operating environment is taken into account.
Depending on the properties of the technologies developed to reach the prior milestones, achievement of this milestone may also require optimization of actuator speed, of MBB lifetime, or of the bonds used to attach new MBBs to the assembler being built.
We have not yet analyzed these suggested product opportunities in any detail, and doing so would require significant effort, but we think that the proposed basic technologies are general enough that many profitable applications could be found, most of which we have never thought of.
One way to look for product opportunities would be to search the literature in various nanoscale-related fields for attempts to make minor nanometer-scale improvements to various materials or processes, where there is an implication that this would have an economic payoff, and then to determine the intended application, and think of ways to make better improvements for the same application.
We organize these potential product ideas by the milestones in section 3 which would need to be achieved before the products could be developed. For each milestone, we assume that all lower numbered milestones have also been achieved. In all cases, further development would be needed which was specific to these products, and which would often require different skills and equipment than the assembler project milestones themselves.
Some of the products listed here would only be practical if large cost reductions could be achieved in synthetic DNA and proteins for use in building blocks.
Note that applications of large crystals of small MBB aggregrates are listed here, whereas applications of large complex MBB aggregates (whether formed into crystals or not) are reserved for the next section.
- "Molecular tips" and receptors for SPMs: Constructs of several MBBs could be designed for attachment to the end of a scanning probe microscope (SPM) tip, to provide it with several different molecular probes (sub-tips, and/or receptors) of known atomic structure and relative position. Such "molecular tip" devices should greatly expand what can be measured and perhaps modified by scanning probe tips, and especially the reliability of such operations, and the tip lifetime. Thus they will be useful for research and perhaps (using tip arrays) for manufacturing. They also have important internal uses for helping Molecubotics do nanoscale research, and possibly for making highly complex assemblies of MBBs "one at a time" (rather than in parallel in a solution.)
- Protein purification: several receptors with affinity to different parts of a protein surface, when arranged around one site with a controlled spatial relationship, chosen by design, would have a much higher affinity for that protein than any individual receptor had. This could be used for exceptionally high-quality purification of that protein. We presume there are markets for purification of proteins meant to be used as pharmaceuticals or in their production, as well as proteins used in biotech research.
There is also an important internal use, since making complex machinery from protein-based MBBs demands very high purity in the base proteins, to avoid incorporating defective MBBs at any single point in the design. A related internal use is to make protein-based MBBs at much lower cost than the present cost of their base proteins, for base proteins which are available in large quantities, but only in a form which is presently too expensive or difficult to purify.
- Small molecule purification: the same idea can be applied to the affinity purification of small molecules, by building "cages" of MBBs which surround sites intended for those molecules with multiple groups with affinity for them. This idea has already proved fruitful when the cages are custom-synthesized large organic molecules; our methodology would provide a simpler systematic way of developing new cages more quickly.
- Artificial zeolites for catalysis: zeolites are useful as specific catalysts because they contain many "cages" of the same internal structure. Another use for "MBB cages" (especially when designed to have specific functional groups on the inside surface of the "cage") would be to ease the design of cages for use as catalysts. Cages synthesized by more traditional methods are already an active area of research in the chemical industry, achieving some success.
- Protein crystallization aids: "cages" made of several MBBs, surrounding a protein of unknown structure (but attached into the cage as if it was also an MBB), could be designed to crystallize, independently of whether the uncaged central protein would crystallize on its own. The resulting crystal structure, if it can be determined to atomic precision by X-ray diffraction, would provide an atomic structure map of the central protein as well. This has markets in pharmaceutical research, since many proteins of interest can't be crystallized directly. Seeman has long had the same application in mind for extended scaffoldings made of DNA. DX molecules extended to work in 3D and to have holes (which Seeman has already proposed) are especially attractive. MBBs other than pure DNA may (or may not) prove easier to use for this.
- MBB kits: the MBBs able to be assembled by our technology could be directly sold, for research or for use in products developed by other companies.
- Biosensors: some existing products consist of complexes of a small number of proteins, such as an antibody and several enzymes, which are able to generate a visible signal (e.g. if the enzymes generate a product which appears blue) at the locations to which the antibodies attach; these are used as probes for the locations of antigens in tissue samples or gels. It seems likely that greater spatial control or higher parts count provided by our methods could be used to optimize some such applications or to develop new ones.
- Protein-based pharmaceuticals might be more reliable if the proteins were arranged in a precisely designed manner, with controlled relative position and orientation.
- Macroscopic filters: Large crystals of small aggregates of MBBs might be used as filters whose pores would be relatively large and have a precise internal structure, able to be functionalized with chosen molecules in a designed way. (For this to be a practical method for large-scale purification, it would probably be necessary to improve MBB lifetime and lower MBB cost, in ways described above.)
- Macroscopic materials with optical, electronic, or electrochemical activity: Large crystals of small aggregates of MBBs, carrying functional groups with optical, electronic, or electrochemical activity, might be able to be developed into improved materials for use in batteries, fuel cells, photovoltaic cells, or other optical, electronic, or electrochemical devices, where the ability to achieve a precise arrangement of the active functional groups is presently a limiting factor, and when the device operation is compatible with the type of MBBs used. Making this practical might require development of non-protein MBBs, and will probably require ways to improve MBB lifetime. It might also require surface modifications to remove undesired activities, or complete removal of water from the final product, if either of these interferes with the intended activity of the device (which seems likely). It would also be necessary to reduce MBB cost in ways described earlier.
- Scaffolding for molecular electronics: several single-molecule electronic devices have been proposed. Seeman and Robinson (1987) have suggested using DNA as a scaffolding for organizing such molecules. Another way would be to build complex aggregates of many MBBs, able to form a 2D or 3D crystal, with the MBBs having surface receptors for specific molecular electronic devices, and also for the ends of molecular wires which would be used to connect them. If this was the first practical method for making complex molecular electronic circuitry, or if it had some significant advantage over other methods, the potential market for assembling high capacity computers and memories would be very large.
Protein lifetime might be an issue in this application (assuming the device-receptors are protein-based), so the lifetime might have to be extended by removing water, crosslinking the proteins, adding polymers to form a non-protein permanent matrix, or by some other means.
### The following subsections are not finished: 4.3 Spin-offs from reaching (3.3) Actuator Milestone 4.4 Spin-offs from reaching (3.4) Simple Machine Milestone 4.5 Spin-offs from reaching (3.5) Grasping Tool Milestone 4.6 Spin-offs from reaching (3.6) Attachment Tool Milestone 4.7 Spin-offs from reaching (3.7) Simulated Assembler Milestone 4.8 Spin-offs from reaching (3.8) The First Biotech Assembler
<section not finished ###>
(proceed with most milestones in parallel)
(note that the best assembler designers are likely to be designers of computer software, or hardware, or mechanical objects, who know some biotech and can talk with its practioners; the Foresight community is a good source of such people. Of course, to develop the components and basic technologies requires people who are already skilled in the application of biotech methods.)