Demonstration of CENTDIST Capability. (a) CENTDIST enhances the Z-score of the AR motif in the AR ChIP-seq data set (LNCaP cell line). The blue bar and red bar show the Z-scores of the AR motif computed using the traditional enrichment method under the window size of 200 bp and the default PWM cut-off (1.32), respectively. The green bars show the Z-score of the AR motif computed by CENTDIST after it optimized different parameters. In the initial stage, the frequency Z-score was calculated using flanking regions at 200 bp as background and default PWM cut-off. In the second stage (window size), CENTDIST finds the best window size to maximize the Z-Score, in which the enrichment window size of AR is changed from 200 to 40 bp. In the third stage (cut-off), CENTDIST finds the best PWM cut-off to maximize the Z-Score, which leads to the flanking region noise level dropping significantly. In the fourth stage, CENTDIST combines the Z-scores of both the frequency graph and the velocity graph, thus further increasing the Z-Score. (b) CENTDIST can repress the Z-score of the false CG-rich motif in the Pol II ChIP-seq data set compared to the traditional overrepresentation methods. All Z-scores are computed exactly as in (a). Since CENTDIST considers the velocity graph of the false CG-rich motif, the combined Z-score of CENTDIST finally drops and is significantly lower than that computed by the traditional enrichment based method. As a side note, this figure also showed that random background can produce quite different results compared to promoter background, which highlights the difficulty of choosing a correct background in existing enrichment based methods.

AP4 is a novel co-TF of AR. (a) ChIP-qPCR of AP4 was performed on 22 randomly selected AR peaks containing AP4 motifs in LNCaP cells before and after 2 h of DHT treatment. (b) Venn diagram depicting the overlap between the ChIP-seq peaks of AR and AP4. (c) AP4 ChIP-seq peak distribution around AR ChIP-seq peak. (d) Association of up-regulated genes with binding sites containing AR + AP4, AR only, or others.

AR motif distribution analysis of our AR ChIP-seq data set. (a) Frequency graph of the AR motif in an AR ChIP-seq data set. (b) Velocity graph of the AR motif in an AR ChIP-seq data set. In each graph, the dotted line partitions the distribution into the enriched region (left region) and the flanking region. The generation of these two graphs can be found in Supplementary Section 1.1.

Co-motif analysis of 13 ES cell TFs using CENTDIST, CEAS and CORE_TF. (a) A comparison of co-motif analysis results using CENTDIST, CORE_TF and CEAS on 13 different ChIP-seq data sets from ES cell. The best setting in each data set for CORE_TF and CEAS were used for comparison. CENTDIST* = CENTDIST algorithm without the inclusion of velocity score. (b) Heat map representing the analysis of 11 ES cell core TFs motif enrichment in 13 ChIP-seq experiments. Every row corresponds to a PWM motif while every column corresponds to a ChIP-seq data set. The color of each entry presents the Z-score (in log scale) of the motifs with respect to the peaks of the ChIP-seq data set. The figure showed that the enhancer motifs are enriched in the enhancer ChIP-seq data sets (top-left gene rectangle) while the promoter motifs are enriched in the promoter ChIP-seq data sets (bottom-right green rectangle).

CENTDIST web interface and program procedure. Users can input or upload ChIP-seq peak locations or the bed format peak-region data, and select the corresponding reference genome and the motif candidates (TRANSFAC, JASPAR or custom database). After submitting the job, the data will automatically be processed according to the CENTDIST analysis pipeline. Specifically, CENTDIST will scan the sequences (±1000 bp around the peaks) and obtain the occurrences of each PWM motif to generate the frequency graph and the velocity graph. Z-score is used to assess the enrichment around peaks for each graph. The center distribution score of each PWM motif is calculated as the sum of the two Z-scores. Finally, CENTDIST outputs a list of TF families ranked by the center distribution scores.