摘要
Transcription factors (TFs) are the core sentinels of gene regulation functioning by binding to highly specific DNA sequences to activate or repress the recruitment of RNA polymerase. The ability to identify transcription factor binding sites (TFBSs) is necessary to understand gene regulation and infer regulatory networks. Despite the fact that bioinformatics tools have been developed for years to improve computational identification of TFBSs, the accurate prediction still remains changeling as DNA motifs recognized by TFs are typically short and often lack obvious patterns. In this study we introduced a new attribute-motif distribution pattern (MDP) to assist in TFBS prediction. MDP was developed using a TF distribution pattern curve generated by analyzing 25 yeast TFs and 37 of their experimentally validated binding motifs, followed by calculating a scoring value to quantify the reliability of each motif prediction. Finally, MDP was tested using another set of 7 TFs with known binding sites to in silico validate the approach. The method was further tested in a non-yeast system using the filamentous fungus Magnaporthe oryzae transcription factor MoCRZ1. We demonstrate superior prediction reranking results using MDP over the commonly used program MEME and the other four predictors. The data showed significant improvements in the ranking of validated TFBS and provides a more sensitive statistics based approach for motif discovery.
Transcription factors (TFs) are the core sentinels of gene regulation functioning by binding to highly specific DNA sequences to activate or repress the recruitment of RNA polymerase. The ability to identify transcription factor binding sites (TFBSs) is necessary to understand gene regulation and infer regulatory networks. Despite the fact that bioinformatics tools have been developed for years to improve computational identification of TFBSs, the accurate prediction still remains changeling as DNA motifs recognized by TFs are typically short and often lack obvious patterns. In this study we introduced a new attribute-motif distribution pattern (MDP) to assist in TFBS prediction. MDP was developed using a TF distribution pattern curve generated by analyzing 25 yeast TFs and 37 of their experimentally validated binding motifs, followed by calculating a scoring value to quantify the reliability of each motif prediction. Finally, MDP was tested using another set of 7 TFs with known binding sites to in silico validate the approach. The method was further tested in a non-yeast system using the filamentous fungus Magnaporthe oryzae transcription factor MoCRZ1. We demonstrate superior prediction reranking results using MDP over the commonly used program MEME and the other four predictors. The data showed significant improvements in the ranking of validated TFBS and provides a more sensitive statistics based approach for motif discovery.
