| Title: | Attractors in Sequence Space: Agent-Based Exploration of MHC I Binding Peptides |
| Author(s): | Jager N; Wisniewska JM; Hiss JA; Freier A; Losch FO; Walden P; Wrede P; Schneider G; |
| Address: | "Chair for Chem- and Bioinformatics, Goethe-University, Siesmayerstr. 70, 60323 Frankfurt, Germany. Institute of Pharmaceutical Sciences, ETH Zurich, Wolfgang-Pauli-Str. 10, 8093 Zurich, Switzerland, N. J. and J. M.W. contributed equally to this study. Charite - Universitatsmedizin Berlin, Tumor Immunology, Chariteplatz 1, D-10117 Berlin, Germany. Charite - Universitatsmedizin Berlin, Institute of Molecular, Biology and Bioinformatics, Arnimallee 22, D-14195 Berlin, Germany. Charite - Universitatsmedizin Berlin, Tumor Immunology, Chariteplatz 1, D-10117 Berlin, Germany. gisbert.schneider@pharma.ethz.ch. Institute of Pharmaceutical Sciences, ETH Zurich, Wolfgang-Pauli-Str. 10, 8093 Zurich, Switzerland, N. J. and J. M.W. contributed equally to this study.. gisbert.schneider@pharma.ethz.ch" |
| ISSN/ISBN: | 1868-1743 (Print) 1868-1743 (Linking) |
| Abstract: | "Ant Colony Optimization (ACO) is a meta-heuristic that utilizes a computational analogue of ant trail pheromones to solve combinatorial optimization problems. The size of the ant colony and the representation of the ants' pheromone trails is unique referring to the given optimization problem. In the present study, we employed ACO to generate novel peptides that stabilize MHC I protein on the plasma membrane of a murine lymphoma cell line. A jury of feedforward neural network classifiers served as fitness function for peptide design by ACO. Bioactive murine MHC I H-2K(b) stabilizing as well as nonstabilizing octapeptides were designed, synthesized and tested. These peptides reveal residue motifs that are relevant for MHC I receptor binding. We demonstrate how the performance of the implemented ACO algorithm depends on the colony size and the size of the search space. The actual peptide design process by ACO constitutes a search path in sequence space that can be visualized as trajectories on a self-organizing map (SOM). By projecting the sequence space on a SOM we visualize the convergence of the different solutions that emerge during the optimization process in sequence space. The SOM representation reveals attractors in sequence space for MHC I binding peptides. The combination of ACO and SOM enables systematic peptide optimization. This technique allows for the rational design of various types of bioactive peptides with minimal experimental effort. Here, we demonstrate its successful application to the design of MHC-I binding and nonbinding peptides which exhibit substantial bioactivity in a cell-based assay" |
| Keywords: | Ant colony optimization Histocompatibility Immunology Machine learning Peptide design Peptides; |
| Notes: | "PubMed-not-MEDLINEJager, Natalie Wisniewska, Joanna M Hiss, Jan A Freier, Anja Losch, Florian O Walden, Peter Wrede, Paul Schneider, Gisbert eng Germany 2010/01/12 Mol Inform. 2010 Jan 12; 29(1-2):65-74. doi: 10.1002/minf.200900008" |
