Current Projects

Scientific Computing for Image-based Systems Biology

It is the central mission of our institute to understand how cells form tissues. How does a cell in the middle of an organ know when to stop growing because the organ is large enough over-all? How does a cell know along which direction to divide to form a well-structured tissue and not just a clump of cells? Cells individually take these decisions based on communication with their neighbours, and on global signals such as mechanical forces and morphogen gradients. Cells hence “compute” when integrating the signals from their environment in order to determine a decision. It is, however, unknown according to which mechanisms and procedures these “computations” unfold.

Our vision is to unravel the “algorithms of tissue formation”, that is the cell-cell communication patterns and intracellular decision making that enable cells to organize into tissues. Cells in a tissue constitute an elaborate stochastic, concurrent computing device. While the components (=molecules) and source code (=genome) are increasingly known, the algorithms enacted remain a frontier for systems biology. It is our long-term aim to reverse-engineer these algorithms for early embryogenesis of C.elegans, D.melanogaster, and D.rerio.

This defines our vision of a “Virtual Tissue”, that is, a computer simulation of a multi-cellular system that combines our current knowledge of biomechanics, molecular pathways, and spatial patterning, enabling us to test if this knowledge is sufficient to explain the growth and form we observe.

Working toward this goal requires a number of computational and theoretical advances, including a modular multi-scale simulation paradigm for biological processes, and interactive real-time microscopy. At the technical level, it also requires advances in high-performance computing and the corresponding programming languages. These are our mid-term goals.

Our past and current research focuses on providing the prerequisites, contributing novel theories and algorithms driven by concrete biological questions. Ultimately, we hope to converge to a mechanistic understanding of how cells communicate, process information and take decisions, within the boundary conditions set by biophysics, and in the systems considered. The biological phenomena considered are flows in the C.elegans egg and the subsequent rounds of cell divisions, as well as zebrafish and drosophila embryogenesis.

We aim to enable the next step for computational biology: simulation of developmental processes in 3D.

Simulating biological processes in complex 3D geometries

Discovering the decision and communication patterns cells execute when arranging into tissues is a reverse-engineering task. In order to see what phenotype and developmental dynamics a cellular “algorithm” produces we need to be able to simulate that “algorithm of tissue formation” in a computer. It is hence our mid-term goal to develop a modular numerical simulation paradigm for biophysical and biological processes in 3D space and time.

We address this by novel hybrid particle-mesh methods that we designed to meet the intricacies of biological systems (i.e., non-linearity, internal activity, complex geometries). Particles are used as collocation points to represent continuous fields. A theory (DC PSE) developed in our group enables consistent approximation of differential operators on any particle function representation. This provides the freedom of distributing the particles arbitrarily, in particular so as to be adapted to the simulated geometries and their temporal dynamics. Further, the particles self-organize according to adjustable interaction potentials, providing unprecedented flexibility to simulate multiple scales of resolution.

So far, we have completed simulation modules for biochemical reaction networks, reaction-diffusion systems, and the first-ever simulation of active biomechanics with the group of Frank Jülicher.

  • Dr. Rajesh Ramaswamy (MPI-PKS)
  • Prof. Dr. Frank Jülicher (MPI-PKS)
  • Dr. George Bourantas (Institute of Fluid Mechanics, Uni Luxembourg)
  • Prof. Dr. Jens H. Walther (Mechanical Engineering, DTU Copenhagen)

Interactive virtual-reality microscopy and image analysis

In addition to predictive forward simulations, reverse-engineering the “algorithms of tissue formation” also requires real-time microscopy, so that we are able to directly overlay simulation results with observations and tune the former using guided black-box optimization algorithms.

A mid-term goal of ours is the development of immersive virtual-reality visualization and hand gesture control for 3D microscopy. Interpretation of user gestures requires real-time image analysis, which is infeasible at the data rates output by modern microscopes. The short-term goal is hence to develop a novel image representation that enables us to process and visualize the data in real time.

We believe that this could become a new standard for the field, in addition to our earlier contributions to image analysis and tracking as available in the MOSAICsuite software package, which is downloaded about 3000 times per month world-wide.

  • Prof. Dr. Eugene W. Myers (MPI-CBG)
  • Dr. Pavel Tomancak (MPI-CBG)
  • Dr. Jan Huisken (MPI-CBG)
  • Prof. PhD Carsten Rother (Computer Vision Group, TU Dresden)
  • Prof. Dr. Stephan Gumhold (Computer Graphics Chair, TU Dresden)
  • Prof. Dr. Raimund Dachselt (Chair for Human/Computer Interaction, TU Dresden)

Parallel high-performance computing for systems biology

Performing simulations, image analysis, and visualization in real time requires parallel high-performance computer resources. In our case, this includes accelerators like graphics cards for visualization, and computer clusters for image analysis and simulations. Combining the two, we develop a standard high-performance environment for particle methods.

To this end, we have been co-developing the Parallel Particle Mesh (PPM) software library since 2005. The PPM Library has enabled some of the world’s biggest and most efficient simulations and has reduced code-development times from years to weeks. In addition to PPM, we also developed a domain-specific programming language for parallel particle methods, PPML. Using PPM and PPML, we are able to implement a fully parallel, scalable particle method for simulation or image processing within a few hours, whereas it took years before to hand-parallelize the code. When used for image processing, PPM and PPML enable real-time image segmentation during microscopy acquisition.

In order to exploit software-development synergies, we participate in the “Dresden Software Synergy Consortium” (Myers, Tomancak, Hiller, Zerial, Sbalzarini), contributing to open-source software projects like Fiji (image processing) and ClearVolume (3D real-time microscopy), and making our methods available as Knime nodes for reusable workflows.

  • Dr. Michael Bussmann (Computational Physics Group, Helmholtz Center Dresden Rossendorf)
  • Prof. Dr. Jeronimo Castrillon (Chair of Compiler Construction, TU Dresden)
  • Prof. Dr. Uwe Assmann (Chair of Software Engineering, TU Dresden)
  • Dr. Pavel Tomancak (MPI-CBG)
  • Prof. Dr. Eugene W. Myers (MPI-CBG)

Mechanics and fluidics of development

Our simulations help unravel the molecular mechanisms of tissue formation across scales, from molecular to organism. If the simulated mechanisms are sufficient to bring about the experimentally observed behavior, we can further use the simulation to determine critical mechanisms that are predicted to be necessary for the behavior. For the latter, black-box optimization and design-centering algorithms are crucial.

A current focus is on understanding the role of fluid flow and active cytoskeletal forces in growth and development. In collaboration with experimental groups, we address these questions on the molecular, cellular, organ, and organism levels. On the molecular level, we work with the Zerial lab to understand how self-organization flows lead to domain formation on endosome membranes that generate the mechanical forces for tethering and fusion of organelles. On the cellular level, we work with Knust lab to understand how this sub-cellular organization establishes and maintains epithelial cell polarity. On the tissue level, we work with the Zerial lab to simulate bile flow in mouse liver in order to study its role during growth and regeneration. Finally, on the organism level we work with the Huisken lab to simulate blood flow in developing zebrafish vasculature in order to analyze the relation between flow mechanics and angiogenesis.

All simulations include either fluid flow or active mechanics. These are particularly challenging phenomena for which new simulation methods are required.

  • Prof. Dr. Marino Zerial (MPI-CBG)
  • Prof. Dr. Elisabeth Knust (MPI-CBG)
  • Prof. Dr. Frank Jülicher (MPI-PKS)
  • Dr. Jan Huisken (MPI-CBG)
  • Dr. Jan Brugues (MPI-CBG)
  • Prof. Dr. Stephan Grill (BIOTEC, TU Dresden)