INESC-ID   Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento em Lisboa
technology from seed


Knowledge Discovery and Bioinformatics
Inesc-ID Lisboa

Mining Protein Structure Data

02/01/2007 - 16:00
02/01/2007 - 17:00

This presentation will show the application of data mining techniques, in particular of machine learning, for discovery of knowledge in a protein database. The main problem we address is the determination whether an amino acid is exposed or buried in a protein for five exposition levels: 2%, 10%, 20%, 25% and 30%. First we introduce the baseline classifier for this problem which, although very simple (only takes into account the amino acid type), already achieves good prediction results. Then we explain how, by making a local PDB database, retrieving DSSP and SCOP data, we build our classifier to improve the baseline prediction. Finally we test and compare several classifiers (Neural Networks, C5.0, CART and Chaid), and parameters that might influence the prediction accuracy. Namely the level of information per amino acid, the SCOP class of the protein and the neighbourhood of the current amino acid (i.e.: the sliding window size).