engleski

Influence of Differences in Experimental Structure Annotations on Accuracy of Structure Prediction of Membrane Proteins

Selection of representative set of proteins to be used for training and validation of bioinformatics method is a key starting step in development of novel reliable methods for prediction of secondary structure of membrane proteins. In the past, and even nowadays, the number of experimentally solved and non-redundant structures of membrane proteins is relatively low, and definitely much lower than the number of unique soluble proteins. Additionally, location of transmembrane segments (either those in alpha or in beta secondary structure) for protein of known structure is not straightforward task. Consequently, the same proteins in different protein databases can have (more or less) different annotations of secondary structure. In this study, starting from information of PDB (Protein Data Bank, http://www.rcsb.org), UNIPROT (http://www.uniprot.org/), and OPM (Orientations of Proteins in Membranes, http://opm.phar.umich.edu) databases, and using ClustalOmega algorithm (http://www.ebi.ac.uk/Tools/msa/clustalo) we created novel representative set of membrane proteins of alpha class at the level of redundancy (in primary structures) below 30%. For this set of membrane proteins we extracted UNIPROT and OPM secondary structure and topology annotations, and all representative sequences were submitted to several representative prediction methods at the http://single.topcons.net/. Accuracies of prediction methods were evaluated in terms of two initial annotations of secondary structure. It was found that significant differences exist beween two experimental structure annotations (i.e. UNIPROT and OPM) and between prediction methods for several membrane proteins.

selection ; representative set ; membrane proteins ; secondary structure ; validation ; prediction method ; redundancy ; transmembrane segments

nije evidentirano