Evaluation of a Method for Predicting Transcription Factors Using Motif-Search Programs

To understand biological phenomena at the molecular level, such as cell-cycle, differentiation, development, proliferation, and so on, it is important to elucidate the mechanism for transcriptional control of each gene, which is involved in their phenomena. Discovery of transcription factors helps us to elucidate the new transcriptional control mechanisms that have not yet been characterized. Although transcription factors can be predicted based on the information derived from homology search using computational tools such as BLAST and FASTA, homology search fails to predict the novel transcription factors that have no significant homology to known transcription factors. Therefore we are evaluating a method for predicting transcription factors based on the existence of local structure such as a DNA-binding domain and/or a transactivation domain, which is often too short to be detected by homology search.