Synthetic biocomputation: the possible and the actual

Computation is defining trait of biological systems and a broad framework that captures the complex adaptive nature of molecules, cells and organisms. Computation is also at the core of the genotype-phenotype mapping, since it provides a natural framework to define function in a self-consistent way. The study of existing biological systems (from signalling cascades to ant colonies or brains) as well as the evolution of synthetic in silico networks performing computations reveals a number of nontrivial patterns of organization, sometimes in clear conflict with standard view of engineering or optimization. In spite of our increasing knowledge, there is a lack of a theoretical framework where computation and its possible forms is integrated within a general picture. Synthetic biology provides a new avenue where engineered molecular circuits can be implemented to perform non-standard computations. Here we review recent advances in the domain of multicellular synthetic computing and suggest a potential morphospace of computational systems including both standard and non-standard approximations. Introduction Computation in nature is a fascinating and yet difficult topic. Biological systems perform computations as they gather information and process it in order to respond to environmental cues. Computation is in fact one formal way of capturing functionality in a well defined fashion (1), (2). Computation has also become a key aspect within the emergent field of synthetic biology (for a recent review, see (19)). This field allows to construct completely new molecular and cellular structures able to perform artificial computations (3). Cells can be engineered in order to behave as autonomous, potentially programmable computing devices. These biocomputing devices would be able to perform complex tasks and designed for a wide range of applications, including bioremediation, food production or biomedicine (4). How to make these systems reusable and scalable is a major problem, but new approaches involving non-standard forms of computing have been able to overcome some key difficulties (5). They define novel ways of computing using living matter and suggest potential scenarios to outline a general framework to unify the landscape of computational structures, both in the natural and artificial realms. Figure 1: Computation occurs in natural systems in many different systems and spanning multiple scales. This include immune networks, social insect colonies, brains or some social amobeae. In order to use computation as a unifying framework where biological complexity and its evolutionary dynamics can be suitably integrated, some formalism is needed. One possibility is to consider classical models of computation. Turing’s formalization of computations in terms of machines with a number of internal states provides a powerful framework where -in principleany potential form of computation could be described (6). The fact that some particular macromolecular systems, such as ribosomes act pretty much as Turing-like nanomachines (reading a ”tape” defined by the messenger RNA, creating an output chain of aminoacids and starting and ending the process by means of detecting given sequences) seems to support this picture. Such avenue has been successfully taken by some researchers (7) proving the viability of making molecular computations close to finite automata. However, as pointed out by Melanie Mitchell (8) there is a range of biological systems, from immune networks to ant colonies or even plants, where computations occur and yet seem to escape from being fully captured by classical, Turing-like formal approaches to computation. The special features shown by information-processing systems in biology have been recognized for decades. Many of them have to do with special ways of treating given computational tasks in a parallel way and using the internal dynamical features characteristic of each system. Task allocation in ants, for example, can be favoured in some cases by means of colony-level oscillations which seem in principle inappropriate for dealing with colony needs. Simple models of ant dynamics based on a neuron-like mapping between ant states and formal neurons have bee very useful in this context. In particular, it has been shown that oscillations actually favour an optimal task fulfilment that is not possible if a constant, average activity level were at work (9), see also (10). Similarly, other properties exhibited by complex biological machines strongly depart from standard engineeringbased principles. One such principle is the robust behavior based on redundancy. Here two identical components of the system making the same function can replace each other in case of failure. Redundancy is thus the intuitive (although sometimes expensive) solution to the problem of failure. However, it has been shown that in many cases (may be in most cases) robust behavior is not obtained from redundant structures. Instead, it seems to be a consequence of so called degeneracy (11), (12), (13). It can be defined as the capacity of elements of a given system that are structurally different to perform the same function or yield the same output. This ubiquitous feature appears to be present in many diferent systems and scales. Modeling in silico evolved circuits performing computations under selection for robust behavior (14) reveal that robustness is achieved through degeneracy, but the underlying mechanistic explanation escapes from our intuition. Degeneracy implies a novel concept beyond standard engineering, suggesting that new forms of thinking might be required. How can we go beyond the limits imposed by real systems, which are the result of evolution and might be difficult to fully characterize? Similarly, how can we test existing theories and try novel ones if they are sometimes difficult to compare with their real counterparts? The field of synthetic biology seems to provide the best scenario for designing novel computational systems in vivo whereas nonstandard forms of computation are used as alternatives to engineering-inspired metaphors. Here we present some of these results and suggest a potential framework to define a space of computational designs that includes existing natural and artificial systems as well as engineered, artificial ones. Logic gates from gene circuits One way of creating synthetic biological circuits performing predefined logic operations is based on engineering genetic a b