Dengue viruses have four serotypes which are members of flaviviruses containing a linear 11 kb single-stranded RNA of positive polarity. In this study we use neural network technology to differentiate different serotypes of dengue viruses. Choice of synonymous codons is not random in many organisms including both prokaryotes and eukaryotes. Patterns and degree of codon usage bias vary not only among different organisms, but also among genes in the same genomes. The idea of relative synonymous codon usage (RSCU) values is to calculate the usage frequency of a particular codon in a codon box. By properly arranging the n virus RSCU values into a matrix format, we are able to adopt the neural network method to distinguish the four serotypes of dengue virus. Our results show that in dengue virus group taxonomically related strains have more similar codon usage patterns than those distantly related strains. The dengue virus codon usage pattern can be used to distinguish different serotypes Keywords-virus; codon usage preference; neural network;I.

No abstract