The mission of the Path is to study emergent behavior and information processing in biological systems and identify principles that underlie biological function and could be beneficial for engineering applications. This is pursued in close cooperation between biologists, engineers, physicists, computer scientists, and mathematicians, combining concepts from the various disciplines. Important challenges are to understand the emergence of robustness, parallelism, convergence, and self-organization in information-processing systems. The goal is to deepen the understanding of the biological systems and to inspire novel engineering applications.

Path Leader: Prof. Marino Zerial

Path Co-Leader: Prof. Marc Timme

Investigators:

• Prof. Dr.-Ing. Dr. h.c. Gerhard Fettweis
• Dr. Benjamin Friedrich (RGL Biological Algorithms Group)
• Prof. Dr. Frank Jülicher
• Prof. Dr.-Ing. habil. Christian Georg Mayr
• Prof. Dr. Ivo Sbalzarini
• Prof. Dr. Marc Timme
• Prof. Dr. Marino Zerial

Previous members:

• Prof. Dr.-Ing. Franz Baader
• Apl. Prof. Dr. habil. Andreas Deutsch
• Dr. Haralampos Hatzikirou
• Prof. Dr. Stefan Siegmund

The goal of the Biological Systems Path is to study emergent behavior of biological systems, such as resilience, parallelism, scalability, self-organization, and energy efficiency, from an information-processing (computation and communication) viewpoint, and to make available this knowledge for innovative engineering applications. We analyze biological examples of information processing where such properties emerge from simple interactions at smaller scales. Examples of such interactions include signals changing from analog to digital (e.g., a cell-fate decision was made), from digital to analog (e.g., turning on a finite set of genes leads to a graded protein concentration), or from stationary to oscillatory (e.g., triggering a cellular oscillation within the ear). How robust (reliable) are these transformations? How is robustness achieved with small numbers of unreliable components? And how can a biological system “solve” complex combinatorial problems with a small number of components? These are some of the key information-processing questions that are addressed in this Path.

Electronic information-processing systems also use different modes of computation (analog/digital) and different scales of implementation. For instance, analog transistors on a chip are used to make digital decisions on the circuit level. In communication systems, digital information is carried by analog signals that are processed by both analog and digital components, before being decoded back to information. Hence, it seems promising to apply some of these concepts to biological information-processing systems, and to compare biological systems with electronic systems in terms of their complexity and (energy) efficiency. Biological systems are among the most complex systems known, and it is possible that universal principles emerge that can provide insight into the increasingly complex engineering systems of the future.

The Bio Path is the outreach and innovation Path of cfaed. It bi-directionally combines two areas where Dresden is strong: microelectronics and the life sciences. As a trans-disciplinary Path, we integrate researchers from both disciplines and allow them to mutually learn from each other. In this, biologists import new concepts and knowledge from engineers and computer scientists, and vice versa. This is fundamentally different from the often-cited “bio-inspired engineering” approaches. Here, the study of the biological system using engineering principles is the driving force, and concepts are imported from the other disciplines. The Biological Systems Path consists of seven interdisciplinary projects, aiming at transferring knowledge bi-directionally between biology and engineering.

Specific biological information-processing systems are analyzed using tools from engineering, computer science, and mathematics. The systems are “reverse engineered” to determine the underlying algorithms and design principles. We quantify the speed, resilience, and efficiency of a broad range of systems from cell, tissue, and developmental biology. The strong emphasis on quantitative data in the biological labs, in combination with the theoretical expertise in the engineering, computer science, and mathematics groups, allows us to obtain definitive answers to key questions concerning biological computing and information processing. Ultimately, we seek to know how closely biological systems approach the theoretical limits set by physics and information theory.

In parallel, new theoretical methods are developed that are applicable to a number of biological problems. These methods include logic-based modeling, communication theory, agent-based simulations, optimization theory, and non-linear dynamics. Because each theoretical approach can be used to analyze different biological problems, the theoretical groups provide important connections between the various projects.

As a bi-directional Path, the engineering principles identified in biological systems are made available to the other Paths in numerous inter-Path collaborations. They may inspire novel ways of thinking about the engineering systems of the future.

The list below provides an overview of all current projects and the participating investigators.

Previous Projects

Project 1
A Logic of Aggregation and Emerging Functionality (Baader)

In biological systems, new functionalities may emerge when small components (e.g., cells) are aggregated into larger entities (e.g., tissue). The purpose of this project is to develop a logic-based modelling language that is expressive enough to describe such phenomena, but still allows effective reasoning. The development of the formalism will be guided by examples found in biological systems, and the resulting formalism evaluated by comparing it to known models from systems biology.

Project 3
Strategies for Controlling Protein Concentrations (Siegmund, Beyer, Gross)

Cellular regulatory processes can be seen as ‘biological computing’. The goal of this project is to understand the processes controlling protein concentrations and to evaluate different regulatory strategies using mathematical models. This work should improve our understanding about which strategies are used in which context in order to meet conflicting optimization goals (e.g. resource efficiency versus speed). Ultimately, this should yield new strategies for implementation in engineered computing systems.

Project 7
Cell fate decision making in multicellular systems: a multiscale approach (Hatzikirou)

Our goal is the profound understanding of cell decision-making in multicellular systems by means of multiscale mathematical modeling. Single cell decision-making has been a major focus of biology. However, in a multicellular system, cell decisions influence and get influenced by its own micro-environment, consisted of homotypic and heterotypic components. This dialogue makes the system dynamics extremely complex.

Replacing the word “cell” by “agent”, we can draw a useful analogy between biological multicellular systems and technical multi-agent ones. The remarkable way that biological system process information can be a source of inspiration for novel engineering ideas.

I.5 project (Robust Synchronisation of Separate Clock Signals on Massively Multi-Core Chips inspired by Biological Systems)

The I.5 project aims at finding novel synchronization strategies for high performance multi-core architectures at low power consumption using concepts developed in the study of biological systems. To this point, the project has obtained interesting new results for networks of mutually coupled electronic oscillators. It has shown how two factors that individually hinder synchronization in such systems will together enable robustly synchronized states. On this basis, novel architectures that allow efficient and robust synchronization are under investigation. The theory is now being tested comparing analytic model predictions to Matlab/Simulink simulations and to experiments with electronic oscillator units, assembled in our group. Compared to prior architectures for globally synchronous operating applications our approach reduces wiring between the oscillators, drops the Master clock and thereby reduces power consumption and production cost. We expect applications in high performance Multiprocessor System-on-Chips (MPSoCs) architectures, distributed antenna arrays, and other large-scale electronic clocking systems communicating by means of time-continuous signals.

I.6 project (Bio-inspired Computing and Optimization Problem Classification)

Optimization problems are ubiquitous in engineering, ranging from parameter estimation to robust design optimization. In practical applications, the function to be optimized is usually not known in mathematical form, but candidate solutions can only be evaluated for quality. Such optimization problems are called "black-box problems". Evaluating the quality of a proposed design or solution in a black-box problem may involve running a computer simulation, performing experimental measurements, or asking an expert.

Due to the practical importance of black-box optimization problems in engineering, hundreds of heuristics and algorithms have been developed, including genetic algorithms, evolution strategies, particle-swarm optimization, simulated annealing, etc. It is a known fact that no algorithm outperforms any other on all possible problems. In practical applications, it is hence of utmost importance to pick a well-suited algorithm for a given problem.

Which algorithm performs well on what type of problems, however, is unknown. It is not even clear how problem "types" or "classes" should be defined such that they predict algorithm performance. This project addresses this shortcoming. It is the goal of this project to develop a recommender system that would suggest a suitable algorithm for a given problem using only the solution evaluations collected. We follow a two-tier strategy: one the one hand, we exploit extensive databases of performance measurements of hundreds of algorithms tested on a standard set of test problems. We use machine-learning methods and tools from computational logic to extract patterns and implications from these data. On the other hand, we seek inspiration in how biology has evolved different optimization strategies for different types of problems. This is motivated by the fact that most black-box heuristics are biologically inspired themselves.

• We have published the idea of a recommender system based on formal concept analysis from logic. The prototype of this system successfully reproduced known recommendations, and suggested new ones.
• We have performed hierarchical bi-clustering analysis of the benchmark database and discovered groups of algorithms that perform similarly well on groups of problems. We currently investigate in what sense these problems are different.
• We generated a catalog of optimization problems and strategies from biology, and are currently relating them to engineering problems and the benchmark problems used in the black-box optimization community
• Applications in other cfAED Paths (parameter estimation in compact models of Carbon Nanotube transistors; robust design centering of flexible thin-film electronics) have shown the potential and challenges of our approach.