Bioinformatics
Institute of Computer Science
University Freiburg
de

Simplified Protein Models

Simplified protein models are coarse grained representations of proteins. They allow for performing large scale studies in silico in order to analyze features of protein-like systems. The main attractivity of such models is that they are still computationally manageable in situations, where the use of more complex models is completely out of computational reach.

Nevertheless for the aim of structure prediction of real proteins, well designed protein models with a reasonable balance of complexity and detail can also provide canditates for further refinement.

Recent studies of protein evolution, thermodynamical, and kinetic aspects of proteins were still strongly restricted to very simple (often 2-dimensional) protein models. Our approaches to protein structure prediction in 3-dimensional protein models have vastly improved the complexity of protein models that can be reasonable used in large scale computational studies.

In this research area, we are interessted in constraint-based techniques for the prediction of optimal protein structures in lattice protein models with strong hydrophobic energy contribution. We solved the protein structure prediction problem for models on the cubic lattice and on the much more complex face-centered cubic lattice. Furthermore, we are interested in all kinds of applications that are enabled by our key technology of structure prediction. For example, we analyze the sequence-structure mapping for a better understanding of protein evolution or construct energy landscapes for finally investigate protein kinetics.

Protein Structure Prediction

Here, we develop methods to predict the spatial structure of simplified proteins in HP-type models on the cubic and face-centered cubic (FCC) lattice. These models are extensions of Ken Dill's HP-model. Using FCC instead of only the three dimensional cubic lattice or even two dimensional lattices is a significant advance, since this lattice can model real proteins rather accurate and lacks the so-called parity problem of cubic lattices.

We are still interested in improvements, either in terms of model complexity or of prediction efficiency.

Current status

We have developed and implemented constrained-based methods for prediction in HP and HPNX models on the cubic and face-centered cubic lattice.

In our implementations, we use advanced constraint-techniques, e.g. general symmetry breaking a la Backofen/Will. CPSP-logo

Available Implementations and Services

Contributing group members

Main Publications

PhD Thesis

Cooperations

Funding

Related Research Topics