Article Preview
Top1. Introduction
Common activities in bioinformatics/cognitive informatics include developing a unified analysis of pattern and organization of biological structures. Developing computational techniques that give insight into these areas is of utmost importance.
The hereditary information for organisms is carried in its genes. Genes are sequences of the polymer DNA which, for our purposes, can be viewed as strings over the alphabet {A,C,G,T}, where each of the four characters corresponds to one of the nucleotide bases that makes up DNA. Individual genes are subsequences of the much larger strings of DNA that comprise the chromosomes of an organism. In addition to specifying the structural information for proteins, genes must be turned on and off at precisely the right time and in the correct tissue in the developing and mature organism. This process is termed as gene regulation and is one of the central problems in modern biology.
The first step in gene regulation is transcription, where the information in a gene is amplified by copying it into RNA, a polymer similar to DNA. Short DNA sequences termed transcription elements, typically of the order of 6-10 base pairs in length, are recognized and bound by sequence-specific binding proteins termed transcription factors to form transcription complexes through protein-protein as well as DNA-protein interactions. Important transcription elements are located immediately preceding the start of genes. More surprisingly, transcription elements are also found thousands of bases upstream, downstream and even within the boundaries of a gene. The transcriptional state of a gene (i.e., its time, tissue and rate of expression) is determined through formation of a “transcription complex” composed of multiple, interacting transcription factors bound to their respective transcription elements. The information needed to specify a transcription factor binding site is not all local to an individual transcription element, but requires interactions with other binding sites through protein-protein interactions to stabilize the complex.
In this paper, we propose to compare promoter sequences by considering the important motifs that are responsible for expression of that particular gene. Some of the available tools that compare promoter sequences are ConReal (Berezikov, Guryev, & Cuppen, 2005) MUMmer (Kurtz, Phillippy, Delcher, Smoot, Shumway, Antonescu, & Salzberg, 2004). Pair wise comparison is possible with these tools and also they do not provide a similarity score. Another class of methods uses prior knowledge of TFBSs to construct the alignments. While ConReal focuses on generating an ordered chain of conserved TFBSs, thus not aligning regions that do not contain them, Siteblast is a BLAST (Michael, Dieterich, & Vingron, 2005)-like heuristic where the TFBS hits are used as seeds. The method of Hallikas et al. (2006) also falls in this category. Here, the sequence of hit pairs is aligned using a scoring scheme that considers clustering of sites, binding affinity and conservation, though the underlying sequences themselves are not aligned. Other approaches like Monkey (Moses, Chiang, Pollard, Iyer, & Eisen, 2004) explicitly take into account evolutionary properties of the TFBSs, but still perform the alignment independent of the annotation step.
The focus of bioinformatics has begun to extend from the identification of genes toward understanding how the expression and regulation of genes is orchestrated in a genomic level. Genes expressed within the same biological context often share promoter modules/frameworks (Fessele, Maier, Zischek, Nelson, & Werner, 2002; Werner, 1999, 2001).