One of our work packages is dedicated to the biophysical and statistical models of cellular context. It is well-known that the presence of the environment of a sub-circuit not only causes extrinsic noise but also memory effects, which means that the dynamics of the subnetwork is affected by its past states via a memory function. In one of our papers, we could give insights into the possible types of memory functions and thereby help to better predict subnetwork dynamics. In another study we tackled the problem of modelling the multi-scale nature of biomolecular reaction networks with hybrid models combining continuous and discrete modeling regimes. In a further paper, we discussed the modeling system of Brownian particles, which can serve as a model for active fluids in cells. An additional research paper contributes to the circuit design task and we could demonstrate an improved performance for a circuit design tool by using e.g. new models to mimic cell-to-cell variability.
To better characterize biological circuits, we developed new statistical inference methods. Statistical inference allows to reconstruct unmeasured molecular states and parameters of a synthetic ciruit from experimental readouts. In two papers we refined the methods to get reliable models out of incomplete data sets and non-exponentially distributed events.
Another work package will deliver methodology to design more robust synthetic circuits. For that we used a minimal model for gene expression, consisting of a switchable promoter together with the resulting messenger RNA, and considered it as a Poisson communication channel. By using filtering theory, we found an exact expression for the mutual information.
To get reliable data for the characterization of biological circuits we want to optimize single-cell experiments. We approach the tasks from two different angles (i) from a computational perspective and (ii) from a technological perspective by improving the experimental hardware and protocol. For the computational perspective we propose moment-based variational inference as a flexible framework for approximate smoothing of latent Markov jump processes (a continuous stochastic process, describing the change of states). The technological perspective was improved by designing and optimizing microfluidics chips for the purpose of circuit characterization on the single-cell level. Cell segmentation is a major bottleneck in extracting quantitative single-cell information from microscopy data. In one of our papers, we present convolutional neural networks trained for multiclass segmenting of individual yeast cells and discerning these from cell-similar microstructures in the deployed microfluidic chips. We give an overview of the datasets recorded for training, validating and testing the networks, as well as a typical use case. We showcase the method’s contribution to segmenting yeast in microstructured environments with a typical synthetic biology application in mind.
Last but not least we experimentally realized cell-free synthetic circuits. We build highly effective AND gates by the combination of small transcriptional activator RNAs (STARs) and toehold switches. To characterize the components and their dynamic range, we used an Escherichia coli cell-free transcription-translation (TX-TL) system. We analyzed a prototype gate in vitro as well as in silico, employing several mathematical methods. On the basis of this analysis all gates were successfully implemented in vivo, offering a dynamic range comparable to the level of the best protein-based circuits.
