Dinucleotide Weight Matrices for Predicting Transcription Factor Binding Sites: Generalizing the Position Weight Matrix

Abstract

BackgroundIdentifying transcription factor binding sites (TFBS) in silico is key in understanding gene regulation. TFBS are string patterns that exhibit some variability, commonly modelled as “position weight matrices” (PWMs). Though convenient, the PWM has significant limitations, in particular the assumed independence of positions within the binding motif; and predictions based on PWMs are usually not very specific to known functional sites. Analysis here on binding sites in yeast suggests that correlation of dinucleotides is not limited to near-neighbours, but can extend over considerable gaps.Methodology/Principal FindingsI describe a straightforward generalization of the PWM model, that considers frequencies of dinucleotides instead of individual nucleotides. Unlike previous efforts, this method considers all dinucleotides within an extended binding region, and does not make an attempt to determine a priori the significance of particular dinucleotide correlations. I describe how to use a “dinucleotide weight matrix” (DWM) to predict binding sites, dealing in particular with the complication that its entries are not independent probabilities. Benchmarks show, for many factors, a dramatic improvement over PWMs in precision of predicting known targets. In most cases, significant further improvement arises by extending the commonly defined “core motifs” by about 10bp on either side. Though this flanking sequence shows no strong motif at the nucleotide level, the predictive power of the dinucleotide model suggests that the “signature” in DNA sequence of protein-binding affinity extends beyond the core protein-DNA contact region.Conclusion/SignificanceWhile computationally more demanding and slower than PWM-based approaches, this dinucleotide method is straightforward, both conceptually and in implementation, and can serve as a basis for future improvements.

Discovery and information-theoretic characterization of transcription factor binding sites that act cooperatively

“…We then determined the corresponding rank sum p-value between the random sets. We found that the p-value of the rank sum test between the DC and DU model fell well beyond the right tail of the random sampling distribution (shown in Figure 2), indicating that the median 30]bp window, which we calculated to have 9.1 bits information relative to the background (uniform distribution of bases), and logo E has the added flanking sites to the 'uncooperative' class. Logo C is the CB motif with 9.6 bits of information relative to the background, which looks similar to the 'uncooperative' class at position 6 due to there being many more sites that prefer A to a T at this position amongst all the Dorsal sites in the network.…”

“…The OR gate is based on the DC detector built from the data set D DC , which contains Dorsal loci from D CB that were tagged with class labels from the optimal spacer window of [0,30]bp with the 5'-CAYATG motif, and similarly, the DU detector is built from the data set D DU , which contains the remaining Dorsal loci from D CB that did not have the Twist sites in the spacer window. .…”

“…For example, if the spacer window is set at [0,30]bp then the DC detector always expects that there is a cooperating site proximal to it, and hence adds w(S, proximal) to the energy E(S) for a given sequence S. All 9-mers that satisfy the constraint E(S) + w(S, proximal) < E c are considered a positive hit (where overlapping 9-mers that satisfy this constraint are screened so that the best scoring 9-mer is considered the positive site).…”

“…We used the same sliding window as in the main text of 30 base pair shifts starting at [0,30]bp, and then incrementing to [31,60]bp etc. We also used the same spacer window as in the main text ([0,30]bp) for the DC detector for making predictions of known Dorsal loci in the CRMs.…”

Automated Incorporation of Pairwise Dependency in Transcription Factor Binding Site Prediction Using Dinucleotide Weight Tensors

“…10 Consequently, a key question in the analysis of gene regulatory networks is to find a 11 proper mathematical representation of the sequence-specificities of TFs. That is, for 12 each TF, we want to determine an energy function E(s) that calculates, for any given 13 DNA segment s, the binding free energy of the TF binding to s. The segment s is 14 generally of fixed length for a given TF, which typically ranges from 6 to 30 base pairs. 15 Although there have been some attempts to use direct structural and biophysical 16 modeling of the sequence-specificity of TFs, e.g.…”