WP6 - IT architecture
Work package 6 is concerned with interfacing the MATCHIT matrix with traditional computing, as well as exploring and exploiting the computational and constructional power of the MATCHIT substrate itself. As such, the work package consists of several thrusts that (a) relate the physical chemistry of chemtainers, DNA tags, and the MEMS device to abstract formalisms of computer science, and (b) link results from computer science and control theory back to the physical device. Figure 6.1 shows the high-level organization of work package 6.
6.1: DNA compiler, container readdressing and MATCHIT language
Partners P1a, P2a, P3b.
We have developed a calculus that allows us to capture system states of the MEMS device (Fellermann and Cardelli 2013). Our calculus uses a grammar akin to brane calculi (Cardelli 2005) to denote distributed arrangements of chemtainers that can be tagged with DNA adresses and can hold arbitrary content including other chemtainers. State transitions can occurr either autonomously (chemical reactions or DNA computations) or be induced via electronic control (chemtainer fusion, chemtainer transport). To control the latter, we have devised a domain specific programming language that consists of few elemental control operations, parallel execution, and potential feedback control.
We have determined the constructive closure of the MatchIT language, i.e. the set of all structures that can be constructed by some sequence of elemental instructions, and we have analyzed how the constructive closure depends on the underlying set of instructions. In particular, we find that a set of eight elemental operations is sufficient to generate any desired system state that can be denoted in the grammar of the MatchIT calculus. Most importantly, the analytical framework that we have developed allows for automatic inference of programs that are able to generate some desired state, i.e. chemical product.
Exponential DNA amplification has been selected as the prefered system of study for the experimental MatchIT work packages. Figure 6.2 shows how DNA amplification can be realized in the MatchIT calculus. For simplicity, we demonstrate DNA amplification by means of a sequential program (left hand side of Figure 6.2), but it is also possible to obtain this amplification reaction in a purely parallel program. A realization of the "split" operation is presented in section 6.4.
6.2: Artificial chemistry simulations of containers
Partners P3a, P4
Programming the MATCHIT matrix to produce a desired target structure with high yield can be a non-trivial control problem. We have implemented and compared different control strategies for the synthesis of complex branched oligomers (e.g. oligosaccharides for glycol-conjugates with medical applications) where potential side reactions impose considerable yield limitations. In MATCHIT, hierarchical compartmentalization (i.e. encapsulation of molecules in chemtainers and positioning of chemtainers in microfluidic environments) is a powerful means to avoid undesired side reactions (Füchslin et al. 2012). Two systems and two control strategies have been studied:
In the first scenario, chemtainers form micro-reactors with specific catalytic properties that can self-assemble into reactor networks which can create a spatially heterogeneous reaction environment with a bias towards specific reaction channels that increase the yield from the reaction (For details of the rather intricate implementation, see Reller 2010). Importantly, this increase is purely due to spatial organization and happens without any kind of interference with the kinetic constants for the reactions.
The second scenario is a cellular automaton model of the MATCHIT matrix as implemented in WP5: The MATCHIT-automaton consists of a tube of cells, each of which is equipped with a set of tags. Chemtainers containing monomers and carrying tags are staged on one side of the tube and are inserted sequentially into the first cell of the tube. After insertion, the chemtainers move through the tube of cells at a speed depending on chemtainer-cell interactions determined by the chemtainer- and cell-tags, respectively. Matching tags induce the chemtainers to stay inside a particular cell for a defined amount of time, allowing for the fusion of two chemtainers if their tags match.
We achieved to realize a "compiler", means an algorithm that analyzes the structure of a branched polymer and generates the description of a specific MATCHIT automaton (In fact, the algorithm we present works for large classes but not all possible branched polymers. For a discussion, see Weyland 2013). A formal description of a branched molecule is translated into "machine code", whereby the term machine code refers to a detailed and complete description of a MATCHIT automaton.
6.3: Physical simulations of interacting containers
Partners P3a, P2a, P1a, P4
P1a has developed physics based molecular simulations of chemtainer formation and dynamics as well as chemtainer-tag interactions. They have developed a physically realistic chemtainer model. The model allows to perform physically realistic Dissipative Particle Dynamics studies of the chemtainer dynamics at the molecular length scale including hydrodynamic effects due to the MEMS environment. The model include the physics of the self-assembly aspects of chemtainers, tag anchoring, and the sequence specific dynamics of the tag addresses. Molecules are represented as strings of soft beads joined by springs. Unfavorable interactions between beads representing e.g. water, oil, surfactants and tags with hydrophobic anchors drive their self-assembly into tagged surfactant covered oil droplets. Droplets are the only chemtainer systems studied so far, since they are computationally simple to study. Beads representing DNA nucleotides have specific interactions modeling Watson-Crick pairing and can be dynamically introduced and broken e.g. by heating the system above the DNA address melting temperature. Importantly, this provides an added layer of physical detail and information, that can form a realistic molecular foundation for describing MATCHIT at the systemic level.
Figure 6.5: DPD simulation of chemtainer relabeling physics (M6.1). The chemtainer here is a surfactant covered oil droplet labeled by 6 tags in the presence of 6 relabeling constructs. Oil is shown as dark green beads. Surfactants head and tail groups are shown as light green and light blue beads, respectively. The tags consist of an address part (small yellow beads) and a hydrophobic anchor (small bright green beads). The relabeling constructs consist of an antiaddress (small orange beads) complimentary to the tag address, and a new address (small red beads). Initially the relabeling constructs are all in solution and the chemtainer type is given by the yellow addresses (left). During the simulation each relabeling construct finds a tag to bond with and in the final state (right) all tags has relabeled with red addresses. Each recognition zone consists of 4 beads with specific pair interactions to ensure stable bonding structure at the reference temperature of the simulation. To emphasize the bonds in the recognition zone (magenta) the sequence beads are shown with a reduced size compared to the oil beads, while in the simulation all interactions has the same range. Note furthermore that periodic boundary conditions apply to the simulations and renderings, e.g. constructs escaping the bottom of the box reenter from the top of the box. Water beads are present but not shown.
Figure 6.6: Six snapshots from two Dissipative Particle Dynamics simulations of freely floating surfactant covered oil droplets. The droplets are labelled by complementary ssDNA tags (blue, red) tethered to the droplets by an anchor. Hybridization bonds are shown as bright yellow. In the simulation shown in the top row, tag complementarity is from tip-to-anchor and tip-to-anchor. The dsDNA tags are observed to a form long lived bridge between the two chemtainers, which eventually leads to fusion. In the simulation shown in the bottom row tag complementarity is tip-to-tip and anchor-to-anchor. In this case, the dsDNA tags pulls the two anchors together, which induces a very fast fusion between the chemtainers.
Partners P2a, P3a
In the case demonstrated a cluster of two different length oligo-nucleotides are to be separated using a hydro-gel.
The hydro-gel is located in a microfluidic channel of 50µm width and 40µm height. The idea behind the controller is to separate both types of oligo-nucleotides to then push them apart once a sufficient distance between both clusters is achieved. Then a negative electrode in between pushes both clusters into opposite directions and allows to process them further using individual algorithms.
In the following a homogeneously filled channel with two types of DNA-oligos (8nt and 30nt) is clustered in a first round. In the second round the DNA-mixture plug is separated along a matrix which is realized via the hydro-gel Pluronic®. After a sufficient separation, which is measured online, the blue painted electrode between the two separated clusters is set to a negative potential and thus separation is finished. In the example shown, the blue cluster is driven back and the red cluster is parked at the same location.
The state-machine used for the experiment follows:
Figure 6.8: Four electrodes are activated and four sensors, which are simple rectangular areas in the camera image, are used to measure the current intensity of the fluorescence signal.
This state-machine is programmed into the controlling software. This software activates the electrode, evaluates the camera image and controls all remaining devices of the whole system. Shown is an excerpt of the setup of the state-machines used for the experiment. Because of the spatially distributed nature several state-machines are connected in a chain-like manner and are cooperatively solving the separation problem.
eval_channel design/fluidic/biopro61_std/M1_B5_SEG.dat control_add autoshift1 "general" r1833 control_add autoshift2 "general" r1834 : : #set_pin r114 high # Activate PIN r114 #set_pin r1618 low # Shielding with negative electrodes #set_pin r718 low # Shielding with negative electrodes auto_shift_8 autoshift1 r1833 r1834 r1835 r1838 r1839 r1840 autoshift2 auto_shift_8 autoshift2 r1834 r1835 r1838 r1839 r1840 r1841 autoshift3 auto_shift_8 autoshift3 r1835 r1838 r1839 r1840 r1841 r1842 autoshift4 : auto_shift_8 autoshift10 r1840 r1839 r1838 r1835 r1834 r1833 autoshift11 auto_shift_8 autoshift11 r1833 r1834 r1835 r1838 r1839 r1840 separ1 separ1 separ1 r1833 r1844 r1845 autoshift12 # This is the separation step auto_shift_8 autoshift12 r1843 r1842 r1841 r1840 r1839 r1838 autoshift13 auto_shift_8 autoshift13 r1842 r1841 r1840 r1839 r1838 r1835 autoshift14 auto_shift_8 autoshift14 r1841 r1840 r1839 r1838 r1835 r1834 autoshift15 auto_shift_8 autoshift15 r1840 r1839 r1838 r1835 r1834 r1833 autoshift1
Newer versions of the software also allow to create moving state-machines which location is then determined by the fluorescence intensity of the moving cluster or droplet or vesicle.
- L. Cardelli, Brane calculi, Lecture Notes in Computer Science 3082:257-280, 2005.
- H. Fellermann, L. Cardelli, Programmatic control of nanoscale bioreactors, in preparation.
- R. M. Füchslin, A. Dzyakanchuk, D. Flumini, H. Hauser, K. J. Hunt, R. H. Luchsinger, B. Reller, S. Scheidegger, R. Walker, "Morphological Computation and Morphological Control: Steps Toward a Formal Theory and Applications", Artificial Life Winter 2013, Vol. 19, No. 1: 9-34.
- Reller, B. (2010). Programmable self-assembling spatially heterogeneous micro-reactors. Master Thesis, Univ. Zürich.
- Mathias S. Weyland, Harold Fellermann, Rudolf M. Füchslin, Daniel Sorek, Doron Lancet, "The MATCHIT automaton: Exploiting Hierarchical Compartmentalization", in preparation.