In contrast to conventional computing devices, artificial cells, like all implementations of embodied computation, exhibit no clear distinction between hard- and software. On one hand, this poses certain restrictions on the feasibility of the most general forms of IT, but on the other hand enables exploiting inherent physical and chemical properties of amphiphilic structures. Evaluating the IT-potential of artificial cells therefore requires a careful choice of the level of abstraction on which the description of processes and interactions between system components is made. Though for a general view, taking account of all molecular details will not be suitable, physical and chemical properties of the hardware have to be included to a degree that enables reflecting the benefits of this real-world embodiment.
The basic components we employ are amphiphilic structures such as vesicles, droplets or micelles. On a more abstract level, they are characterized by the fact that they form two-compartment systems. One of these compartments is the interior, in case of vesicles an aqueous solution, in case of droplets a lipid environment. The other compartment is the boundary, in case of vesicles a membrane. Both of these compartments are able of providing a suitable environment for a large variety of reactions. Two facts are essential: First the topological relation between them makes the boundary an interface between environment and interior. Second, the boundary, especially in case of vesicles, forms a two-dimensional fluid which itself is a reaction environment for molecules embedded into the membrane. Noteworthy, there is a non-trivial surface-to-volume relation between size and mutual interaction of chemical components of the interior and boundary.
Vesicles and droplets interact with each other by exchange of signals and material. Though nature realized sophisticated ways for target-specific communication, in the systems we investigate, interaction depends on the relative position of the compartments: Transfer between them is driven by diffusion.
We distinguish four different types of IT-relevant systems:
1. Systems emulating conventional IT, e.g. membrane computing, DNA-computing, tagging systems and the like.
2. Externally programmable systems.
3. Morphological computation for process control.
4. Evolving systems.
Following the PACE-objectives, we focussed on last three types. Externally programmable systems do not necessarily, in the strict sense of the term, process information, though they are, by virtue of their programmability in all cases, part of an IT-process. A paradigmatic example that served us as a guiding vision is the creation of a programmable platform for the implementation of complex chemical processes. Thereby, chemical process chains are subdivided into a number of steps, and those will be realized in reaction vessels given by vesicles or droplets. These vesicles may have different content and exchange of matter happens mostly between direct neighbours. The latter implies that the total process is determined by the spatial arrangement of the vesicles. Using appropriate key-and-lock molecules, we could show that vesicles can be selectively coupled to each other or to a substrate. The linkers don't depend on the content and can be attached to "raw" (not yet coupled but already loaded) vesicles at any time. The choice of the linkers determines the spatial arrangement of the vesicles: based on an appropriate library of vesicles with different chemical content, a large multitude of possible processes can be "programmed" and realized. Parallel process control is a typical IT-problem. In PACE, we investigated fundamental processes such as sorting of objects or establishing non-Fickian gradients in diffusion systems controlled only by morphology-dependent interactions between the system components, either with themselves or with the environment. We regard this as an application of the paradigm of morphological computation. In a general context, we characterize morphological computation by contrasting two of its most relevant features with conventional computing.
1. In a conventional computation, the relevant aspects of the physical reality underlying the process to control are mapped onto a digital representation. It is this representation, which then is processed further by employing some sort of algorithm. Finally, the outcome of this algorithm is translated back in actions of physical systems. Conventional computation therefore poses a number of non-trivial requirements to a control process. First, the mappings between reality and representation have to be implemented. Second, the problem to be solved employing algorithms on the digital representation of the physical situation has to be efficiently tractable. We emphasize that tractability is often related to the properties of the representation and not the problem itself: It is much harder to calculate the dynamics of a non-linear system than a linear one, though for the nature this distinction is irrelevant, both processes just take place following the laws of physics. It is a key aspect of morphological computation to allow other than digital representations of a problem and therefore to exploit nature's indifference towards, say, problems of numerical analysis.
2. The number of variables employed is of high relevance for a conventional computation. Nature in contrast is inherently parallel. This does not only mean that e.g. many particle processes pose no speed or communication problem as in conventional computation, but also that soft objects, characterized by more than the six parameters of a rigid body, can be treated without loss of efficiency. Finally, evolving systems do not only process, but produce information. For evolution to take place, a number of conditions need to be fulfilled. First, the systems have to members of a combinatorially rich family of different types of realizations. Second, they must be reproducible in a manner that the type of the offspring is correlated with the parent's type, means inheritance has to take place. And third, the must be a type-function relation, such that selection can take effect. In PACE, McMullin and co-workers discussed the effects of natural replication schemes in depth. Relying on technologies developed by Eggenberger and Hadorn, Füchslin (in cooperation with Eggenberger, Hadorn, Hanzcyc and Shaw) sketched a novel replication scheme termed "configurational inheritance" which is presently still under investigation but directly takes profit of the selective linking of vesicular structures. Software On an abstract level, the implementation of functionality by replicating and interacting compartment (vesicle/droplet) systems exhibits features of functional as well as imperative programming. On the level of an individual compartment, the implementation of a complex reaction scheme given by a network of pathways can be understood as a form of imperative programming, with the designer using recognition mechanisms to guarantee the proper sequential order of process threads. As soon as different process steps are distributed over several types of compartments, a process comprised of a sequence of actions is not only (or even not at all) the result of a careful a molecular addressing scheme but (also) emerges from the spatial organization of the individual functions. If these spatial structures are even self-organized, implementing a complex reaction network then follows the paradigm of functional programming. In practice, and in order to upheld a sort global process management, synchronization issues start to play an important role.
In PACE, important results for systems of artificial replicating cells have been obtained by Serra and the Sole group. Exploiting the spatial arrangement of chemically activated and interacting compartments requires some sort of control of their placement as a necessary and already far reaching partial result. Eggenberger and Hadorn demonstrated the feasibility of achieving this by selective adhesion molecules (see above).
1. Systems emulating conventional IT, e.g. membrane computing, DNA-computing, tagging systems and the like.
2. Externally programmable systems.
3. Morphological computation for process control.
4. Evolving systems.
Following the PACE-objectives, we focussed on last three types. Externally programmable systems do not necessarily, in the strict sense of the term, process information, though they are, by virtue of their programmability in all cases, part of an IT-process. A paradigmatic example that served us as a guiding vision is the creation of a programmable platform for the implementation of complex chemical processes. Thereby, chemical process chains are subdivided into a number of steps, and those will be realized in reaction vessels given by vesicles or droplets. These vesicles may have different content and exchange of matter happens mostly between direct neighbours. The latter implies that the total process is determined by the spatial arrangement of the vesicles. Using appropriate key-and-lock molecules, we could show that vesicles can be selectively coupled to each other or to a substrate. The linkers don't depend on the content and can be attached to "raw" (not yet coupled but already loaded) vesicles at any time. The choice of the linkers determines the spatial arrangement of the vesicles: based on an appropriate library of vesicles with different chemical content, a large multitude of possible processes can be "programmed" and realized. Parallel process control is a typical IT-problem. In PACE, we investigated fundamental processes such as sorting of objects or establishing non-Fickian gradients in diffusion systems controlled only by morphology-dependent interactions between the system components, either with themselves or with the environment. We regard this as an application of the paradigm of morphological computation. In a general context, we characterize morphological computation by contrasting two of its most relevant features with conventional computing.
1. In a conventional computation, the relevant aspects of the physical reality underlying the process to control are mapped onto a digital representation. It is this representation, which then is processed further by employing some sort of algorithm. Finally, the outcome of this algorithm is translated back in actions of physical systems. Conventional computation therefore poses a number of non-trivial requirements to a control process. First, the mappings between reality and representation have to be implemented. Second, the problem to be solved employing algorithms on the digital representation of the physical situation has to be efficiently tractable. We emphasize that tractability is often related to the properties of the representation and not the problem itself: It is much harder to calculate the dynamics of a non-linear system than a linear one, though for the nature this distinction is irrelevant, both processes just take place following the laws of physics. It is a key aspect of morphological computation to allow other than digital representations of a problem and therefore to exploit nature's indifference towards, say, problems of numerical analysis.
2. The number of variables employed is of high relevance for a conventional computation. Nature in contrast is inherently parallel. This does not only mean that e.g. many particle processes pose no speed or communication problem as in conventional computation, but also that soft objects, characterized by more than the six parameters of a rigid body, can be treated without loss of efficiency. Finally, evolving systems do not only process, but produce information. For evolution to take place, a number of conditions need to be fulfilled. First, the systems have to members of a combinatorially rich family of different types of realizations. Second, they must be reproducible in a manner that the type of the offspring is correlated with the parent's type, means inheritance has to take place. And third, the must be a type-function relation, such that selection can take effect. In PACE, McMullin and co-workers discussed the effects of natural replication schemes in depth. Relying on technologies developed by Eggenberger and Hadorn, Füchslin (in cooperation with Eggenberger, Hadorn, Hanzcyc and Shaw) sketched a novel replication scheme termed "configurational inheritance" which is presently still under investigation but directly takes profit of the selective linking of vesicular structures. Software On an abstract level, the implementation of functionality by replicating and interacting compartment (vesicle/droplet) systems exhibits features of functional as well as imperative programming. On the level of an individual compartment, the implementation of a complex reaction scheme given by a network of pathways can be understood as a form of imperative programming, with the designer using recognition mechanisms to guarantee the proper sequential order of process threads. As soon as different process steps are distributed over several types of compartments, a process comprised of a sequence of actions is not only (or even not at all) the result of a careful a molecular addressing scheme but (also) emerges from the spatial organization of the individual functions. If these spatial structures are even self-organized, implementing a complex reaction network then follows the paradigm of functional programming. In practice, and in order to upheld a sort global process management, synchronization issues start to play an important role.
In PACE, important results for systems of artificial replicating cells have been obtained by Serra and the Sole group. Exploiting the spatial arrangement of chemically activated and interacting compartments requires some sort of control of their placement as a necessary and already far reaching partial result. Eggenberger and Hadorn demonstrated the feasibility of achieving this by selective adhesion molecules (see above).