Naslov
Nucleotide coding of amino acid polarity and protein structure

Autori
Štambuk, Nikola ; Konjevoda, Paško ; Gotovac, Nikola

Izvornik
Acta Universitatis Debreceniensis de Ludovico Kossuth Nominatae, Seria Physica et Chimica (0567-7947) 34-35 (2002); 171-188

Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni

Ključne riječi
protein; folding nucleotide; amino acid; sequence; machine learning; classification tree; genetic code; dipeptide

Sažetak
New computational algorithm for the prediction of alpha and beta protein folding types from the nucleotide and amino acid sequences is described. The method is based on the binary coding of physicochemical nucleotide and amino acid properties. The algorithm enables quick, simple and accurate prediction of alpha and beta protein folds on a personal computer by means of few binary patterns of coded physicochemical properties. It was tested by means of machine learning SMO (sequential minimal optimisation) classifier and classification trees, on a dataset of 140 dissimilar protein folds. Depending on the method of testing, the overall classification accuracy was 91.43%-100% and the tenfold cross-validation result of the procedure was 83.57%-85%. Genetic code randomisation analysis based on 100 000 different codes tested for the protein fold prediction quality indicated that: a) there is a very low chance of p = 2.7 x 1 0^-4 that a better code then the one specified by the binary coding algorithm is randomly produced, b) dipeptides represent basic protein units with respect to the genetic code defining of alpha and beta protein folding types.

Izvorni jezik
Engleski

Znanstvena područja
Temeljne medicinske znanosti

POVEZANOST RADA