engleski

PREDICTIONS OF MEMBRANE PROTEIN TOPOLOGY

For integral membrane proteins with known amino acid sequence the very first task is to predict the sequence location of transmembrane segments and their orientation with respect to membrane surfaces. Finding ends of transmembrane helices is not an easy goal even for small number of membrane proteins whose crystal structure has been obtained. When this is accomplished the information about likely membrane associated secondary structure has been reduced to the two-dimensional cartoon labeled as membrane protein topology. The quality of this information is very important in pharmacology and medicine in order to determine potential location of drug binding pockets close to external or internal membrane surface. Predicting the function also requires accurate knowledge of membrane protein topology, which is often better conserved than its sequence. We have developed a set of high quality algorithms for accomplishing the above mentioned goals. The essential procedure during development of these tools has been intelligent extraction of biological information from carefully selected non-redundant data sets and objective performance testing by using independent data sets of integral membrane proteins. Our algorithm SPLIT is currently among best in the world for predicting sequence location of transmembrane helices (http://split.pmfst.hr/split/4/) when tested on independent dataset available at the TMH_benchmark server. It also offers a good platform for accurate prediction of pore segments in cationic channels, and the possibility to examine the performance of any amino acid scale in predicting membrane associated structures. The web portal SPLIT, available on-line for sequence analysis, is fast and simple so that students and research workers from more than 100 universities are regularly using it. The most recent improvement in SPLIT, introduced to reduce the number of false positive predictions in signal peptide region, was achieved by utilization of the knowledge-based database of signal sequences. Also, with optimal choice of amino acid scales we achieved better discrimination between soluble and membrane proteins. In a future we hope to combine the Croatian high quality tools SPLIT and GARLIC (University of Osijek – http://garlic.mefos.hr/garlic ) in order to create novel algorithm for tertiary structure prediction of membrane proteins.

membrane proteins; prediction; topology; transmembrane helices

nije evidentirano