Entwicklung von Methoden des Maschinellen Lernens zur funktionellen Charakterisierung von Peroxisomen

• Antragsteller: Dr. Thomas Lingner
• Fachliche Zuordnung: Bioinformatik und Theoretische Biologie
• Förderung: Förderung von 2011 bis 2014
• Projektkennung: Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 189920290

Sequence-based prediction of the precise function of particular cell compartments such as peroxisomes is of growing importance, e.g. to elucidate the response of crop plants to environmental stress or to identify drug targets in pathogens by pathway analysis. The aim of the proposed project is to develop machine learning algorithms for functional characterization of whole cell organelles on the example of peroxisomes in different organisms. For this purpose, approaches to peroxisomal protein prediction and prediction of functional properties and metabolic pathways based on protein domains will be combined. Firstly, known peroxisomal proteins from species of different kingdoms will be collected and used to substantially improve the accuracy of an existing machine learning approach to the prediction of peroxisome-targeted proteins by integration of different sequence features. Secondly, an existing method for microbial phenotype prediction based on the protein domain content of an organism shall be adapted to the problem of deriving functional properties of a peroxisome based on its protein inventory. Furthermore, existing methods for prediction of metabolic pathways from protein sequence sets will be extended and applied to identify relevant peroxisome-specific pathways. Finally, the approach is planned to be extended to metagenomics data such that functional properties of peroxisomes in fungal communities can be described quantitatively.

• DFG-Verfahren: Sachbeihilfen