Sequential Modeling for Identifying Gene Locations in Human Genome

We consider several sequential processing algorithms for identifying genes in human DNA, based on detecting CpG islands. The algorithms are designed to capture the underlying statistical structure in a DNA sequence. Sequential processing using a Markov model and a hidden Markov model are shown to identify most CpG islands in annotated (marked) DNA subsequences in publicly available DNA data sets. We also consider a wavelet-based hidden Markov tree (HMT). In the context of the HMT, we address design of adaptive wavelets matched to CpG islands, this effected via lifting and geneticalgorithm optimization. DNA is comprised of a sequence of units called nucleotides [1]: adenine (A), cytosine (C), guanine (G), and thyamine (T). In the human genome, a C nucleotide is generally modified chemically by methylation if followed by a G. Methyl-C mutates into a T with a high probability. The methylation process is suppressed in localized segments of the genome, often at the “start” regions of many genes. These regions, characterized by a higher concentration of C-G dinucleotides than elsewhere, are called CpG islands (“C precedes G”). Our objective is to produce models for distinguishing variable-length CpG islands from the rest of the DNA sequence. We consider the following algorithms: a