Previous Frame Next Frame
  • Welcome

    We are an interdisciplinary junior research group at Freie Universität Berlin developing methods for analyzing biological systems, ultimately allowing for new clinical diagnostics tools. Proteomics is a relatively young discipline dealing with the complicated protein mixture in biological systems, for example the human body. Changes in the concentration of some protein species can often be directly linked to diseases, such as cancer. The Mecical Bioinformatics Group develops new algorithms capable of identifying changes in the very high dimensional proteome data we are getting from clinical partners. In particular, we are interested in tiny signals that are usually hidden in the noisy part of the data and not detectable by current approaches.

    Model 1
  • Welcome

    We are an interdisciplinary junior research group at Freie Universität Berlin developing methods for analyzing biological systems, ultimately allowing for new clinical diagnostics tools. Proteomics is a relatively young discipline dealing with the complicated protein mixture in biological systems, for example the human body. Changes in the concentration of some protein species can often be directly linked to diseases, such as cancer. The Mecical Bioinformatics Group develops new algorithms capable of identifying changes in the very high dimensional proteome data we are getting from clinical partners. In particular, we are interested in tiny signals that are usually hidden in the noisy part of the data and not detectable by current approaches.

    Model 2
  • Welcome

    We are an interdisciplinary junior research group at Freie Universität Berlin developing methods for analyzing biological systems, ultimately allowing for new clinical diagnostics tools. Proteomics is a relatively young discipline dealing with the complicated protein mixture in biological systems, for example the human body. Changes in the concentration of some protein species can often be directly linked to diseases, such as cancer. The Mecical Bioinformatics Group develops new algorithms capable of identifying changes in the very high dimensional proteome data we are getting from clinical partners. In particular, we are interested in tiny signals that are usually hidden in the noisy part of the data and not detectable by current approaches.

Bio Patterns

Follow this link to browse our Bio Patterns databases.

The Team

Meet the team and see what we are doing.

Our Software

Here you can find information about our software libraray.

Are you looking for some interesting data-sets?

Click here

Data-driven modeling of cellular processes and beyond

This research is carried out in the framework of Matheon supported by Einstein Foundation Berlin.

   
Project heads: Tim Conrad (FU/ZIB), Stefan Klus (FU), Christof Schütte (FU/ZIB)
Staff: Dr. Wei Zhang (ZIB)

Project Background

Cellular processes are governed by diffusion, transport, and interactions of its constituents. For many processes the spatial inhomogeneity of cells is of secondary importance; modelling such processes means finding appropriate kinetic models of the underlying cellular reaction networks (CRNs). The availability of such models is key to many areas of the life sciences ranging from computational biology to system medicine and is essential for understanding the fundamentals of cellular behavior, its malfunction under external stress and its restoration by regenerative interventions.

The goal

The project aims at developing DDMI into a technique that can be applied to observational trajectory data of cellular processes. The main goal is to identify CRN models directly based on available time-resolved process data without the need to identify the reactions that have to be included in advance. A similar attempt has been made before: In in-silico biology, for example,  evolutionary algorithms were used to construct optimal models via "random mutations" of the underlying CRNs. Unfortunately, these approaches were not very successful, mainly due to the lack of discrimination power between inappropriate models and artefacts resulting from ill-conditioned parameter identification. The approach proposed herein does \emph{not} face the same problem: In DDMI, parameters are "automatically" identified as the leading expansion coefficients in the ansatz space. That is, the ill-conditioned inverse problem of the parameter estimation step is avoided.

DDMI is of particular importance for modelling the effect of therapeutic interventions in cellular dynamics (like in the context of regenerative medicine). For such processes, almost no reliable models are available. DDMI based on time-resolved models of the system including the external influences could be of enormous help.

In particular, we aim at identifying approximate eigenfunctions of transfer operators associated with high-dimensional systems from simulation or measurement data for:

  • Dimensionality reduction.
  • Detection of metastable sets.
  • Separation of time-scales.
  • System identification.
 
Publications