Prediction Of Transcription Factor Binding Sites And Combinatorial Regulation Of Eukaryotes

Identification of transcription factor binding sites is an important step towards the understanding of the transcription regulation. Reliable prediction of transcription factor binding sites can be used to identify the target genes of transcription factors and infer the relationship between the positions of the binding sites and regulation activity of transcription factors. We have developed an approach to identify over-represented transcription factor binding sites in a group of related sequences. Over-represented motifs identified in a group of related regulatory sequences may play an important role in transcription regulation of the genes. By comparing the group of related sequences and several set of background sequences, our algorithm can identify over-represented motifs in the related sequences. Our algorithm was based on the Positional Weight Matrix(PWMs) of known transcription factor binding sites, the prediction result can be directly linked with the transcription factors. The testing of the algorithm indicates a relatively high specificity of the result. With more PWMs become available, our algorithm can present more useful information for the understanding of the transcription regulation. A web-based transcription factor binding prediction server was also launched.Genes are usually controlled by several transcription factors. The understanding of the combinatorial regulation of transcription factors is important for full understanding of the transcription regulatory network. The investigation on combinatorial regulation requires the understanding of the logic relationship between transcription factors and their regulated genes. Currently, no high through-put methods can be used to measure the activity of different transcription factors under different experimental conditions. We developed a probabilistic model to modeling the activity states of transcription modules with microarray data under different experimental conditions. We defined the module as a collection of genes with similar expression behaviors. Hence, genes already clustered with microarray data can be modelled as a transcription module. The transcription activities of known clusters under different conditions were inferred with our probabilistic model. With the inferred activities, the relationships between the motifs presented in the regulatory region of genes and the activity of the genes under givencondition were analyzed. Motifs and motif-motif pairs with significant correlation with the activity state of genes under given conditions were identified. Such motifs and combination of motifs provide useful information for further research in combinatorial regulation. We also developed an algorithm to identify modules with our probabilistic model. The algorithm can identify genes belonging to the same module from a set of candidate genes. The transcription activities of those genes were also inferred. The activities of several modules are highly correlated with the corresponding transcription factors. Such modules can be used to infer the activities of transcription factors.