Hidden markob models for computational gene recognition

Abstract

Genes, some long molecules of DNA, store the control codes for all the activities of life; and scientists are giving huge efforts to tind out the genetic codes in the cells of different living beings, especially of humans. Because of the huge volume of the databases containing the genomes of various species, computational gene recognition tools have become essential for discovery and analysis of the genes. The genes constitute only little portions of the genomic DNA sequences, and are interleaved by long non-coding intergenic regions. There are interlcaving of coding and non-coding regions within the genes too. The problem of gene recognition is to identify gene in the huge volume of DNA sequence, and also to identify the coding and non-coding regions inside the gene. This thesis describes a new and simple Hidden Markov Model based system, namely HMMSplice for recognition of donor and acceptor splice sites in a genomic DNA sequence. Since identification of splice sites extracts the coding exons and non-coding introns in a gene and thus, completely reveal the structure of a gene, this system provides substantial aid for recognizing genes in un-annotated DNA sequences. Hidden Markov Models provide a precise probabilistic method for modeling sequence of discrete data, and therefore seem to be a natural solution for analyzing various sites in DNA sequences. Separate HMMs ii)r donor and acceptor splice sites have been designed for HMMSplice. They are trained and tested with real data and the results of the experiments have been discussed. Since complete understanding of the biological process that recognizes and utilizes genes to synthesize proteins, is essential to develop as well as lhlderstand a gene recognition system, a comprehensive discussion on protein synthesis is provided Tbe features of spilce sites that arc considered in development of the models, are discussed in detaiL