A non-independent energy-based multiple sequence alignment improves prediction of transcription factor binding sites

A novel method for improved accuracy of transcription factor binding site prediction

Identifying transcription factor (TF) binding sites (TFBSs) is important in the computational inference of gene regulation. Widely used computational methods of TFBS prediction based on position weight matrices (PWMs) usually have high false positive rates. Moreover, computational studies of transcription regulation in eukaryotes frequently require numerous PWM models of TFBSs due to a large number of TFs involved. To overcome these problems we developed DRAF, a novel method for TFBS prediction that requires only 14 prediction models for 232 human TFs, while at the same time significantly improves prediction accuracy. DRAF models use more features than PWM models, as they combine information from TFBS sequences and physicochemical properties of TF DNA-binding domains into machine learning models. Evaluation of DRAF on 98 human ChIP-seq datasets shows on average 1.54-, 1.96- and 5.19-fold reduction of false positives at the same sensitivities compared to models from HOCOMOCO, TRANSFAC and DeepBind, respectively. This observation suggests that one can efficiently replace the PWM models for TFBS prediction by a small number of DRAF models that significantly improve prediction accuracy. The DRAF method is implemented in a web tool and in a stand-alone software freely available at http://cbrc.kaust.edu.sa/DRAF.

Knowledge-based three-body potential for transcription factor binding site prediction

A structure-based statistical potential is developed for transcription factor binding site (TFBS) prediction. Besides the direct contact between amino acids from TFs and DNA bases, the authors also considered the influence of the neighbouring base. This three-body potential showed better discriminate powers than the two-body potential. They validate the performance of the potential in TFBS identification, binding energy prediction and binding mutation prediction.

TRX-LOGOS - a graphical tool to demonstrate DNA information content dependent upon backbone dynamics in addition to base sequence

BACKGROUND

It is now widely-accepted that DNA sequences defining DNA-protein interactions functionally depend upon local biophysical features of DNA backbone that are important in defining sites of binding interaction in the genome (e.g. DNA shape, charge and intrinsic dynamics). However, these physical features of DNA polymer are not directly apparent when analyzing and viewing Shannon information content calculated at single nucleobases in a traditional sequence logo plot. Thus, sequence logos plots are severely limited in that they convey no explicit information regarding the structural dynamics of DNA backbone, a feature often critical to binding specificity.

SOFTWARE AND IMPLEMENTATION

We present TRX-LOGOS, an R software package and Perl wrapper code that interfaces the JASPAR database for computational regulatory genomics. TRX-LOGOS extends the traditional sequence logo plot to include Shannon information content calculated with regard to the dinucleotide-based BI-BII conformation shifts in phosphate linkages on the DNA backbone, thereby adding a visual measure of intrinsic DNA flexibility that can be critical for many DNA-protein interactions. TRX-LOGOS is available as an R graphics module offered at both SourceForge and as a download supplement at this journal.

RESULTS

To demonstrate the general utility of TRX logo plots, we first calculated the information content for 416 Saccharomyces cerevisiae transcription factor binding sites functionally confirmed in the Yeastract database and matched to previously published yeast genomic alignments. We discovered that flanking regions contain significantly elevated information content at phosphate linkages than can be observed at nucleobases. We also examined broader transcription factor classifications defined by the JASPAR database, and discovered that many general signatures of transcription factor binding are locally more information rich at the level of DNA backbone dynamics than nucleobase sequence. We used TRX-logos in combination with MEGA 6.0 software for molecular evolutionary genetics analysis to visually compare the human Forkhead box/FOX protein evolution to its binding site evolution. We also compared the DNA binding signatures of human TP53 tumor suppressor determined by two different laboratory methods (SELEX and ChIP-seq). Further analysis of the entire yeast genome, center aligned at the start codon, also revealed a distinct sequence-independent 3 bp periodic pattern in information content, present only in coding region, and perhaps indicative of the non-random organization of the genetic code.

CONCLUSION

TRX-LOGOS is useful in any situation in which important information content in DNA can be better visualized at the positions of phosphate linkages (i.e. dinucleotides) where the dynamic properties of the DNA backbone functions to facilitate DNA-protein interaction.