Application of Kernel Method to Reveal Subtypes of TF Binding Motifs

Abstract

AbstractTranscription factor binding sites often contain several subtypes of sequences that follow not just one but several different patterns. We developed a novel sensitive method based on kernel estimations that is able to reveal subtypes of TF binding sites. The developed method produces patterns in form of positional weight matrices for the individual subtypes and has been tested on simulated data and compared with several other methods of pattern discovery (Gibbs sampling, MEME, CONSENSUS, MULTIPROFILER and PROJECTION). The kernel method showed the best performance in terms of how close the revealed weight matrices are to the original ones. We applied the Kernel method to several TFs including nuclear receptors and ligand-activated transcription factors AhR. The revealed patterns were applied to analyze gene expression data. In promoters of differentially expressed genes we found specific combinations of different types of TF binding patterns that correlate with the level of up or down regulation.KeywordsGene Expression DataGibbs SamplingWeight MatriceKernel MethodWeight KernelThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.