DNA familial binding profiles made easy: comparison of various motif alignment and clustering strategies

Abstract

Transcription factor (TF) proteins recognize a small number of DNA sequences with high specificity and control the expression of neighbouring genes. The evolution of TF binding preference has been the subject of a number of recent studies, in which generalized binding profiles have been introduced and used to improve the prediction of new target sites. Generalized profiles are generated by aligning and merging the individual profiles of related TFs. However, the distance metrics and alignment algorithms used to compare the binding profiles have not yet been fully explored or optimized. As a result, binding profiles depend on TF structural information and sometimes may ignore important distinctions between subfamilies. Prediction of the identity or the structural class of a protein that binds to a given DNA pattern will enhance the analysis of microarray and ChIP-chip data where frequently multiple putative targets of usually unknown TFs are predicted. Various comparison metrics and alignment algorithms are evaluated (a total of 105 combinations). We find that local alignments are generally better than global alignments at detecting eukaryotic DNA motif similarities, especially when combined with the sum of squared distances or Pearson's correlation coefficient comparison metrics. In addition, multiple-alignment strategies for binding profiles and tree-building methods are tested for their efficiency in constructing generalized binding models. A new method for automatic determination of the optimal number of clusters is developed and applied in the construction of a new set of familial binding profiles which improves upon TF classification accuracy. A software tool, STAMP, is developed to host all tested methods and make them publicly available. This work provides a high quality reference set of familial binding profiles and the first comprehensive platform for analysis of DNA profiles. Detecting similarities between DNA motifs is a key step in the comparative study of transcriptional regulation, and the work presented here will form the basis for tool and method development for future transcriptional modeling studies.

Figure 1. Illustration of Familial Binding Profile Construction

In this example, the binding motifs for four bZIP–CREB transcription factors are aligned in a multiple-motif alignment. The generalized familial binding profiles correspond to the weighted average of the individual profiles.

Figure 2. Distribution of the Observed Scores of Column-to-Column Comparisons for the Five Main Similarity Metrics

Columns are obtained from the TRANSFAC database [17]. The ALLR_LL distribution is identical to ALLR for every point ≥2 (unpublished data). Comparison of the JASPAR motif columns yielded similar results.

Figure 3. Performance of the Five Main Similarity Metrics in Discriminating between Columns Sampled from Dirichlet Distributions around Information Content I and a Background Distribution

The plot shows the positive predictive rate for an FDR of 1% as a function of the information content.

Figure 4. Average Homogeneity of Families Represented at Each Tree Node as a Factor of the Growth of the Tree

Six scoring metrics and two different tree-building methods are tested with ungapped Smith–Waterman alignments.

Figure 5. The Tree Resulting from a UPGMA Tree Construction of Ten JASPAR Families (71 Motifs Total) Using the PCC Scoring Metric and Smith–Waterman (Ungapped) Alignment Method

The red line represents the level at which the CH _log metric estimates the optimal number of data clusters on the tree.

Figure 6. The Behaviour of the Calinski and Harabasz–Based Log-Metric (CH _log) for the Tree in Figure 5 as the Number of Clusters (g) Is Varied

The value of g = 17 produces a global minimum in the value of CH _log.

Figure 7. The Tree Resulting from a UPGMA Tree Construction of 12 JASPAR Families (79 Motifs Total) Using the PCC Scoring Metric and Smith–Waterman (Ungapped) Alignment Method

This tree includes the two zinc-finger families (GATA and DOF).

Figure 8. Optimal Number of Clusters of the 71 JASPAR Motifs, According to Our Method

PCC with Smith–Waterman ungapped alignment was used as a scoring function. Examples of protein–DNA complexes are provided for comparison.

Figure 9. Similarity between the HMG and Forkhead Motifs

These families are grouped together on the HMG/Forkhead Group I cluster (Figure 8).