engleski

Symbolic coding of amino acid and nucleotide properties

A large body of literature relates physical and chemical properties of the amino acids and protein folding. However, relatively little is known about the relationships of the codons and secondary protein structure. Nucleotide strings based protein folding prediction is important because the Genome Project has resulted in a large number of gene sequences that code for different proteins, of often unknown structure and function We present a new computational algorithm for the prediction of all alpha- and all beta-protein fold classes from the nucleotide and amino acid sequences. The method is based on the binary and error-control coding of nucleotide and amino acid physicochemical properties. It enables quick, simple and accurate prediction of all alpha- and all beta-protein folds on a personal computer. The analysis of the prediction accuracy and its cross-validations by means of the machine learning SMO classifier and classification trees has confirmed the validity of the procedure. We also investigated how encoding physicochemical parameters of the nucleotides and amino acids influences the structure of protein. Genetic code randomisation analysis with respect to the distinguishing of all alpha- and all beta-protein fold classes indicated that: a) there is a very low chance that a better code than the one specified by the nature is randomly produced, b) basic protein units with respect to the genetic coding of alpha- and beta-protein fold classes are not monomers (single amino acids) but dipeptides and tripeptides.

nucleotide; amino acid; protein; structure; genetic code

nije evidentirano