Official websites use. Share sensitive information only on official, secure websites. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Characterizing the binding preferences of transcription factors TFs in different cell types and conditions is key to understand how they orchestrate gene expression. Here, we develop TFscope, a machine learning approach that identifies sequence features explaining the binding differences observed between two ChIP-seq experiments targeting either the same TF in two conditions or two TFs with similar motifs paralogous TFs.
TFscope systematically investigates differences in the core motif, nucleotide environment and co-factor motifs, and provides the contribution of each key feature in the two experiments. Gene expression programming is the primary mechanism that controls the cellular phenotype and function. PWMs can be used to compute binding affinities and identify potential binding sites in genomes. However, contrary to bacterial DBDs which recognize sequences that often have sufficient information content to target particular genomic positions, most eukaryotic DBDs recognize short binding motifs around 10 bp that are not sufficient for specific targeting in the usually large e.
These studies showed that most TFs only associate with a small subset of their potential genomic sites in vivo [ 57 ] and that the binding sites of a given TF often vary substantially between cell types and conditions [ 51 ]. Furthermore, as the number of DBD families in a genome is small relative to the number of TFs, TF paralogs from the same DBD family often share very similar binding motifs, yet they usually show distinct binding sites in vivo [ 3 , 27 , 33 , 49 ].
On the other hand, several studies have revealed that a substantial number of the in vivo binding sites lack an obvious match with the known binding motif of the target TF [ 33 , 57 ]. At this point, it is important to emphasize the strong links that exist between TF binding and histone marks [ 18 ]. However, it remains unclear whether these chromatin states are a cause or a consequence of TF binding [ 26 ]. Moreover, recent approaches based on machine learning, and specifically convolutional neural networks CNNs , have shown that transcription factor binding, but also gene expression, histone modifications, and DNase I-hypersensitive sites, can be predicted just from DNA sequences, often with surprisingly high accuracy [ 2 , 30 , 42 , 56 , 59 , 63 ].
