The overall objective ofWP5 is to develop novel programming methodologies for IT coprocessors to chemical cells. This involves a novel combination of multiple levels of software. In part the tools exist already, but their combination and integration is novel. The central idea is that one first characterize a physical model of the chemical system (including its self-assembly and reactive properties), feed this into a component model description describing small pieces of the chemical microprocessor circuitry, integrating this into a real time UML style model description. Then one takes the parameterized model components, with the states of electrodes and the chemical composition and starting configurations as input variables, and connects them with transport models (which are also pre-simulated) into a model of local region of the electronic chemical microprocessor. These modules are then tested and verified (and possibly corrected) against their performance in the achievement of target objectives (e.g. moving molecules from A and concentrating them at B, as indicated by real time fluorescence imaging) using the feedback of information and the model to predict the required electrode control sequences. This form of combinatorial control has been tested already in the PACE project, but needs to be extended to handle the novel range of modules required for ECCell. Once several modules have been so defined, the combinatorial algorithm space for electrode control will be decomposed into functional modules (e.g. EOF control, pH control, concentration, travelling wave separation and then an appropriate genetic encoding of the space of control programs will be formulated). The information associated with a modular electrode control program will be associated with the ECCell location (electronic genome) and may be optimized and even coevolved with the chemical system. The investigation of this hardware-software coevolution in a real hybrid system showing the full functionality of a cell, with shared information processing between the electronic system and the chemical system will be the core goal of this program development initiative.
Simulation of multiscale and hybrid systems
Multiscale simulations of self-assembly and chemical kinetic phenomena are necessary to reach the right space and timescales, inaccessible to molecular dynamics. Following experience with many coarse-grained simulation tools, including lattice and off-lattice models, from Lattice-Boltzmann to Ginzburg-Landau, in ECCell we shall initially concentrate on one core method, combining it with recent improvements in the efficiency of stochastic chemical kinetics beyond the Gillespie algorithm. As the central physical simulation method, we will employ dissipative particle dynamics (DPD), a coarse-grained particle based method that conserves momentum and is therefore especially suited for investigations of the dynamics of extended objects composed of large number of individual entities. DPD combines three types of forces: conservative interactions, determining the macroscopic dynamics of extend objects, and dissipative and random forces that integrate the effects of molecular motion on faster timescales in a thermodynamically consistent way. The BioMIP group has extended classical DPD by including chemical reactions and by enabling an efficient self-organized treatment of supramolecular structures via macroscopically parameterizable multipolar potentials . The motivation behind this extension is the idea that a DPD-particle’s dipole moment defines a local direction and can be understood as a surface or line element that can be used to build up extended curved lower-dimensional objects such as membranes and networks embedded in space. This extension is compatible with other recent extensions improving the efficiency of DPD. BioMIP has already connected this simulation with detailed microfluidic component geometry descriptions, enabling the development of a modular component simulation tool as required in this project.
Molecular Information Processing
Molecular information processing is a complex and rapidly growing area ranging from nanoscale circuitry in organic materials, through nanobiophotonics (e.g. bacteriorhodopsin) to DNA computing and membrane computing. Whereas one extreme deals with finding efficient molecular devices for classical information processing, and another deals with novel paradigms for solving hard combinatorial problems, it is now clear that one main technical innovative value of molecular information processing is the fact that it is embedded (or immersed) in the molecular world. There it can mediate complex parallel and real time processes of analysis, synthesis and interprocess communication, for which the computational or interface resources via conventional computers would be prohibitive. Thus Adleman’s solution of SAT problems with DNA computers have given way to DNA self-assembly [1] and DNA mediation machines which through sequences of hybridisation reactions can process molecular signals (linking molecular events to amplification processes in programmable ways.
Programming languages for controlling adaptive and hybrid systems
There is a vast literature concerning modeling of dynamical systems which in the beginning was restricted to continuous systems and later extended to hybrid-systems incorporating also discrete events or automata. The stability of hybrid systems has been investigated and special programming languages designed. Many programming languages concerning biological systems have been developed in the areas of Molecular Computing, or with formal specifications as in π-calculus, Petri-nets and in the study of hybrid systems. These hybrid systems also have been applied to modeling biological systems. A promising concurrent language is Hybrid cc but UML-RT (unified modeling language, real time ) diagrams promise to provide a more tractable programming access point. This has already been discussed in the context of Systems Biology with SBML (systems-biology-markup-language) and its spatial extensions as the communication platform of cellular models. Also the developments in the membrane molecular computing paradigm deserve further consideration as a basis for a programming language for artificial cells and dynamically compartmentalized programmed chemical systems. What is missing form these formulations are collective self-organization and self-assembly processes and the notion of instruction scale feedback control loops (or ubiquitous regulation) as a basis for adaptivity. We intend to bridge this gap in ECCell. Similarly, developments in genetic algorithms and programming, classifier systems and in particular self-assembling microprocessors provide a powerful additional starting point for exploring embedded real time adaptive and evolutionary programming.
