Short Abstract: We seek to tightly regulate transgene expression in biotech crops. To that end, we want to identify transcription factor binding site (TFBS) sequence motifs associated with useful expression patterns. Here we present POWRS, a novel algorithm for discerning such motifs from publicly available, genome-wide sequence and expression data. POWRS overcomes two important limitations of conventional motif finders. First, many algorithms ignore motif positioning. POWRS uses a simple but robust statistical model for motif positioning that actually increases its ability to detect weakly enriched motifs. Second, many algorithms tend to over-generalize, merging similar but distinct TFBS into a single, uninformative motif. POWRS uses a discrete alternative to position weight matrices (PWMs) that better resists over-generalization, thereby identifying true consensus TFBS motifs that are useful for engineering gene expression.
We first validate POWRS on a published benchmark of human TFBS and miRNAs, showing that it performs at least as well as other widely-used programs. Then, we apply POWRS to identify the preferred position and sequence of three motifs associated with constitutive high expression in the model plant Arabidopsis thaliana. Finally, we validate these predictions by systematically mutating endogenous Arabidopsis promoter sequences. In most cases, mutations in the identified motifs decrease expression significantly more than other nearby mutations, suggesting that the sequences identified by POWRS do in fact play an important role in establishing the desired expression pattern.

Short Abstract: Analyses using high-throughput techniques (e.g. tiling microarrays and RNA-seq) revealed a much higher complexity of transcriptomes than originally expected. This includes a high number of long non-coding RNAs (lncRNAs) whose functions are still not known in most cases. Affymetrix GeneChip Exon 1.0 ST Arrays (exon arrays) were designed to measure the expression of each exon of protein coding genes. In addition, they contain an even larger number of probes for ESTs and putative genes from gene predictions. However, those probes were completely discarded in most exon array analyses. We utilized these undervalued probes to generate a custom annotation for exon arrays that focuses on measuring the expression of lncRNAs. By re-analyzing existing exon array data sets, we were able to identify tissue- and developmental-specific lncRNAs as well as lncRNAs that undergo alternative splicing events.
Besides the custom annotations, we provide an easy-to-use web interface for biologists without extensive knowledge about programming. Our web interface processes exon array CEL files based on our custom annotation to reveal expression changes of lncRNAs. An integrated genome browser, visualizations of signals and various other features allow a detailed inspection of the analyzed data. The web server is freely available at http://noncoder.mpi-bn.mpg.de.

Short Abstract: To monitor eukaryotic mRNA metabolism, we developed comparative Dynamic Transcriptome Analysis (cDTA). cDTA provides absolute rates of mRNA synthesis and decay in Saccharomyces cerevisiae (Sc) cells with the use of Schizosaccharomyces pombe (Sp) as internal standard. cDTA uses non-perturbing metabolic labeling that supersedes conventional methods for mRNA turnover analysis. cDTA reveals that Sc and Sp transcripts that encode orthologous proteins have similar synthesis rates, whereas decay rates are five-fold lower in Sp, resulting in similar mRNA concentrations despite the larger Sp cell volume. cDTA of Sc mutants reveals that a eukaryote can buffer mRNA levels. Impairing transcription with a point mutation in RNA polymerase (Pol) II causes decreased mRNA synthesis rates as expected, but also decreased decay rates. Impairing mRNA degradation by deleting deadenylase subunits of the Ccr4-Not complex causes decreased decay rates as expected, but also decreased synthesis rates. Extended kinetic modeling reveals mutual feedback between mRNA synthesis and degradation that may be achieved by a factor that inhibits synthesis and enhances degradation. cDTA is provided with a statistical methodology and all required bioinformatics steps that allow the accurate absolute quantification and comparison of RNA turnover. cDTA can be applied to reveal rate changes for all kinds of perturbations, e.g. in knock-out or point mutation strains, as responses to stress stimuli or in small molecule interfering assays like treatments through miRNA or siRNA inhibitors. The cDTA approach is in principle applicable to virtually every organism.

Short Abstract: Human iNOS (hiNOS) gene expression is regulated by transcriptional and post-transcriptional mechanisms. The purpose of this study was to determine whether specific microRNA (miRNA) directly regulate hiNOS gene expression. Sequence analysis of the 496 bp hiNOS 3’-untranslated region (3’-UTR) revealed five putative miR-939 binding sites. The hiNOS 3’-UTR conferred significant post-transcriptional blockade of luciferase activity in human A549, HCT8, and Hela cells. The hiNOS 3’-UTR also exerted basal and cytokine-stimulated post-transcriptional repression in an orientation-dependent manner. Functional studies demonstrated that transfection of miR-939 into primary human hepatocytes (HC) significantly inhibited cytokine-induced NO synthesis in a dose-dependent manner that was abrogated by a specific miR-939 inhibitor. MiR-939 (but not other miRNAs) abolished cytokine-stimulated hiNOS protein in human HC, but had no effect on hiNOS mRNA levels. Site-directed mutagenesis of miR-939 bindings sites at +99 or +112 bp in the hiNOS 3’-UTR increased reporter gene expression. Furthermore, intact miR-939 binding sites at +99 or +112 positions were required for post-transcriptional suppression by miR-939. Cytokine stimulation directly increased miR-939 levels in human HC. Transfection of miR-939 inhibitor (anti-sense miR-939) enhanced cytokine-induced hiNOS protein and increased NO synthesis in vitro in human HC. Taken together, these data identify that miR-939 directly regulates hiNOS gene expression by binding in the 3’-UTR to produce a translational blockade. These findings suggest dual regulation of iNOS gene expression where cytokines induce iNOS transcription, and also increase miR-939 leading to translational inhibition in a check-and-balance system.

Short Abstract: Transcription of genes in eukaryotes is regulated by clusters of cis-regulatory elements (CREs) called cis-regulatory modules (CRMs) through their interactions with transcription factors (TF). However, current understanding of CRMs in any eukaryotes remains rudimentary due to the high-cost of experimental characterization and computational prediction complexity. With the development of next-gen sequencing technologies, Chip-seq has become a cost-effective routine to locate binding regions of target TFs genome-wide, and a large volume of dataset has been generated. We assume that although each Chip-seq dataset are enriched of the target TF, they should also indicate the locations of possible CRMs involving the TF. Therefore, CRMs can be predicted by integrating information in a large number of Chip-seq datasets. We tested this hypothesis using 145 Chip-seq datasets for 58 TFs in the model organism D. melanogaster, and found that: (1) Binding regions in this small set of TFs are extensively overlapping within a short peak range, suggesting the co-occurrence of CREs in the sequences. (2) In addition to target TF motifs, multiple other statistically significant motifs are found in each dataset. (3) A large portion of predicted CREs from different datasets are overlapping with one another, indicating that one motif exists in multiple datasets, forms CRMs with others. By assuming that over-presented co-occurrence patterns of the predicted motifs in these datasets are likely form CRMs, we developed a graph clustering method to predict CRMs. With rapidly accumulation of ChIP-seq datasets, this method should provide more comprehensive and accurate prediction of CRMs in any organism.

Short Abstract: We performed a meta-analysis of data generated as part of the ENCODE project to provide a comprehensive map of transcription start sites (TSSs) across the human genome. Our analysis included the integration of 57 datasets reporting TSS location using CAGE, 7 datasets profiling H3K4me3 marks, and 85 datasets profiling DNaseI hypersentivity. This data was combined using machine learning techniques to provide a map of all TSS sites across the human genome.
To demonstrate the utility of the TSS map, we analysed two GRO-SEQ datasets (profiling the 5’ end of actively transcribed regions) in MCF7 and LnCAP cell lines after estrogen and testosterone treatment, respectively. The TSS map enabled comprehensive expression analysis of all TSSs in the genome. Furthermore, gene set enrichment analysis (GSEA) showed improved pathway identification using our TSS map compared to the typically used refGene or AceView TSS annotations. In addition, the application of our TSS map to RNA-SEQ data analysis also showed improved performance when compared to standard gene annotations.
Incorporating our TSS map into RNA-SEQ analysis pipelines yields increased sensitivity and unbiased analysis of the expression of all TSSs across the genome. Initial analyses show increased estimated rates of anti-sense transcription. The TSS map was also used to analyze the relationship between transcription factor binding and transcript regulation through integration of RNA-SEQ profiling and ChIP-SEQ profiling in the same cell lines. Initial results demonstrate shared and distinct relationships between TF binding and TSSs for the TFs ER, AR, STAT1, p73 and p53.

Short Abstract: Whole-genome and reduced representation bisulfite sequencing (BS and RRBS) provide a powerful way to investigate DNA methylation levels at single nucleotide resolution. However, substantial computational work is required to interpret the raw results from such experiments. In bisulfite converted reads from BS-seq data, unmethylated cytosines are converted to thymines while methylated cytosines remain unchanged. Traditional aligners are not able to map these types of reads. Our previously published algorithm, BS Seeker, was explicitly developed for this task. In this project we developed BS Seeker 2, an enhanced pipeline for aligning and analyzing both BS and RRBS data. It is integrated with the most updated versions of SOAP2, Bowtie and Bowtie2 for both single-end and pair-end mapping. Aligner options are fully customizable via the command line interface. Coupled with Bowtie2, BS Seeker 2 is able to perform local alignments, which may resolve better incomplete adaptor trimming as well as other potential 3’ and 5’ sequence contamination. Also, BS Seeker 2 utilizes gapped alignments which improves the overall quality of results. Our tool is extensively tested by using synthetic reads as well as mapping against large genomes. In comparison to current aligners, BS Seeker 2 is faster and more accurate. The source code is publicly available and is integrated within the Galaxy web platform where scientists can use it as a part of their complete analysis pipeline.

Short Abstract: Chromatin immunoprecipitation followed by sequencing (ChIP-seq) is an effective method for genome-wide profiling of DNA-binding proteins. We conducted ChIP-seq experiments for more than 60 different transcription factors in M. tuberculosis and developed a computational pipeline for the analysis of the data obtained. We were able to identify transcription factor binding events at high resolution and to assign potential motifs to previously uncharacterized transcription factors. Positions of binding events were used to find potential target genes for transcription factors.
We analyzed the complementary overexpression data to estimate the probability of a transcription factor regulating a target gene through a binding site knowing the target gene expression range. We found that about 25% of all predicted binding sites were validated with available expression data. Higher percentage of intergenic binding sites was significant compared to genic binding sites. About 40% of significant binding sites were located around 5’-ends of their target genes; however, binding sites located in coding regions were also validated better than the random model.
We used the binding information for orthologous genes in M. tuberculosis and M. smegmatis (Rv3574 and MSMEG_6042, Rv3133c and MSMEG_5244, Rv0081 and MSMEG_6457, Rv3249c and MSMEG_1842, Rv0757 and MSMEG_5872) to analyze evolution of regulation in Actinomycetales. We studied the patterns of binding motif appearances and disappearances on the Actinomycetales phylogenetic tree and correlate it with available expression and functional data. We showed that the evolutionary story of the regulation depends on transcription factor conservation and role in the organism (e.g., comparison of kstR and dosR).

Short Abstract: The modern high-throughput sequencing methods such provide massive amounts of DNA-positioned data. This data is often represented as a function (e.g. coverage) of the DNA coordinate. The genomewide or chromosomewide correlations of this data from different sources provide information about functional biological relation of the investigated features, e.g. trancription and histone modification. The task to compute the correlation was successfully solved for interval annotations (http://genometricorr.sourceforge.net/) as well as for general case (Ramsey et.al. 2010, Bickel et.al. 2010, Bickel et.al. 2009)
Here we present the method to reveal correlation between two biological features along with estimation of statistical significance of the correlation. The method is based on fast Fourier transform (FFT). The orthogonal properties of Fourier harmonics allow to calculate the correlation, as well as to make permutations and thus compute the p-value quite fast.

References
1. Ramsey SA, Knijnenburg TA, Kennedy KA, Zak DE, Gilchrist M, Gold ES, Johnson CD, Lampano AE, Litvak V, Navarro G, Stolyar T, Aderem A, Shmulevich I: Genome-wide histone acetylation data improve prediction of mammalian transcription factor binding sites.
Bioinformatics 2010, 26:2071-2075
2. Bickel P, Boley N, Brown J, Huang H, Zhang N (2010) Subsampling methods for genomic inference. Annals of Applied Statistics 4: 1660–1697
3. Peter J. Bickel, James B. Brown, Haiyan Huang and Qunhua Li (2009) An overview of recent developments in genomics and associated statistical methods. Phil. Trans. R. Soc. A 2009 367, 4313-4337

Short Abstract: To analyze gene regulatory networks, the DNA/RNA binding affinities of proteins and non-coding RNAs are crucial. These are often deduced from sequences enriched in factor binding sites. The binding site motifs are best described by position weight matrices (PWMs). Up to now, the statistical significance of PWMs has been computed using time-consuming sampling approaches. Therefore, the PWM-based methods instead all maximize a likelihood.

Here we present XXmotif (eXhaustive evaluation of matriX motifs), the first PWM-based motif discovery method that directly optimizes the statistical significance of enrichment. It computes 100,000s of single-site P-values for thousands of candidate PWMs during the refinement, using an efficient branch-and-bound algorithm that calculates exact single-site P-values with an eighth-order background model. This approach allows us to naturally combine P-values for motif enrichment, conservation, and localization.

When tested on ChIP-chip/seq, miRNA knock-down, and co-expression data sets from yeast and metazoans, XXmotif outperforms state-of-the-art tools, both in numbers of correctly identified motifs and in the qualities of PWMs. In human core promoters, XXmotif reports in a single run most previously described and five novel motifs sharply peaked around the transcription start site, among them an Initiator motif similar to the one in flies.

Short Abstract: Bisulfite treatment of DNA followed by next-generation sequencing (Bisulfite-seq) has become a popular method for profiling DNA methylation epigenomic state, yet no accurate method exists for determining genotype polymorphisms in bisulfite-seq data. Single-nucleotide polymorphisms (SNPs) can result in inaccurate or missing methylation calls. We have developed a bisulfite-seq module for the Genome Analysis Toolkit (GATK) called Bis-SNP which uses bayesian inference with either manually specified or automatically estimated prior probabilities to determine genotypes and methylation levels simultaneously. We use validation data for 1 million array-based genotype calls to confirm the accuracy of Bis-SNP on Bisulfite-seq data, and investigate the genomic coverage necessary to detect most heterozygous SNPs in bisulfite-seq data.
We have adapted Bis-SNP to analyze data based on a new bisulfite-based method, Nucleosome Occupancy and Methylation (NOMe-seq). NOMe-seq allows for the simultaneous determination of DNA methylation and nucleosome
occupancy by using a methyltransferase enzyme that footprints nucleosome positions at GpC dinucleotides and leaves CpG methylation state unchanged. Accurate genotyping of heterozygous SNPs using Bis-SNP allows us to investigate allele-specificity of combined methylation and nucleosomal patterns. Specifically, we investigated human fibroblast cells to understand the relationship between nucleosome depleted regions (NDRs) and DNA methylation at gene promoters, and found a large number of promoters had combined epigenomic patterns specific to a single allele. We also investigated these relationships at other elements with NDRs and positioned nucleosomes, including enhancers and CTCF insulators. Bis-SNP is freely available and is a powerful tool for identifying allele-specific epigenetic regulation using either standard bisulfite sequencing or NOMe-seq.

Short Abstract: The function of a cell depends on the correct choice of the genes to be expressed. Gene expression is mediated by regulatory proteins, which recognize and bind specific DNA regions. Understanding how binding occurs in vivo is an essential step to understand gene regulation. A global view of the binding network can be obtained experimentally using chromatin immuno-precipitation (ChIP) followed by sequencing (seq). Currently, most analysis of ChIPseq focus on how to identify regions corresponding to true binding events. However, little has been explored to understand binding inside those regions. In this research, we developed an approach to decode the binding patterns, with special interest in regions with multiple sites. We started developing a parametric model to represent the binding signal. ChIPseq coverage is consequence of DNA fragments purified during immuno-precipitation, and we assume two possibilities: single binding for cases in which immuno-precipitated DNA contains only one binding event, and double binding, for case with two. We incorporated our model into a ChIPseq data analysis pipeline which uses sequence motifs to constrain search space and increase the resolution of binding site detection. This pipeline is able to detect multiple sites inside the same region and outperforms current methods. It also shows high-reproducibility in both M. tuberculosis and humans ChIPseq data. Modeling double binding events increases sensitivity and specificity of site detection when compared to the case considering only single binding. Finally, we estimate cooperative binding from ChIPseq by exploring the secondary terms provided by the double binding signal.

Short Abstract: Allele-specific gene expression is an important source of phenotypic variation. Studying the underlying regulatory mechanisms of allele-specific gene expression helps link genetic and epigenetic variations with phenotypic changes. Recent advances in sequencing technologies and the availability of genome-wide experimental data from large international consortia like ENCODE and the 1000 Genomes Project have enabled us to investigate allele-specific transcriptional regulation on an unprecedented scale. Using the rich datasets provided by the consortia, we conducted an integrated analysis to identify allele-specific protein-DNA interactions, histone modifications, and DNase I hypersensitive sites throughout the human genome, and correlated them with allele-specific gene expression. We uncovered novel allele-specific co-regulations among TFs, and observed well-correlated allele-specific TF binding, epigenetic changes, and gene expression. A survey of the genetic variations at binding sites for more than 30 transcription factors (TFs) showed that occurrence of heterozygous single nucleotide polymorphisms (SNPs) within TF binding sites frequently leads to allele-specific TF binding (50-80% frequency in two-thirds of the TFs), providing a strong genetic basis for allele-specific gene regulation. In summary, we identified tens of thousands of functional SNPs in non-coding regulatory regions, which greatly expanded the current catalog of allele-specific regulatory sites, and our approach approves to be an effective strategy for uncovering functional variants in the human genome.

Short Abstract: Identification of splicing regulatory elements (SREs) deserves special attention because these cis-acting short sequences are vital parts of splicing code. The fact that a variety of other biological signals cooperatively govern the splicing pattern indicates the necessity of developing novel tools to incorporate information from multiple sources to improve splicing factor binding sites prediction. Under this context, we proposed a Varying Effect Regression for Splicing Elements (VERSE) to discover intronic SREs in the proximity of exon junctions by integrating other biological features. As a result, 1562 intronic SREs were identified in 16 human tissues, many of which overlapped with experimentally verified binding motifs for several well-known splicing factors, including FOX-1, PTB, hnRNP A/B, hnRNP F/H, and so on. The discovered tissue, region, and conservation preferences of the putative motifs demonstrate that splice site selection is a complicated process that needs subtle and delicate regulation. VERSE may serve as a powerful tool to not only discover the splicing elements by incorporating additional informative signals but also precisely quantify their varying contribution under different biological context.

Short Abstract: Exquisitely controlled during the development of the organisms and to vital functions, the transcription of the information encoded within the DNA to an RNA molecule has as the protagonist the enzyme RNA polymerase (RNAP). The development of single-molecule techniques, as magnetic and optical tweezers, atomic-force microscopy and single-molecule fluorescence, advanced our understanding of the phenomenon. It was noted that the RNAP does not maintain the same transcription rate during the polymerization, "pausing" at specific sites. Theoretical models have been proposed to explain and predict the occurrence of these pauses. However, experiments showed that the elongation of the same open reading frame (ORF) is different if more than one RNAP molecule initiates from the same promoter. We proposed a stochastic sequence-dependent model based on the Gillespie algorithm, able to identify collisions between RNAP and predict their cooperative behavior. In our approach, when we have collisions between the enzymes, one RNAP exerts a net force
of 25pN on the leading molecule. In our model the transcription is 1.16 to 1.60 times more efficient than a model where interactions do not occur. This result occurs because the
colisions reduce the dwell times at the
pause sites. These results are consistent with the experiments, that shows that some
sequences have a transcription rate increase up to 2.10 times due this efect.

Short Abstract: The amount of sequencing data output generated from Next Generation Sequencing platforms has dramatically increased year after year. As the output has increased, the time requirements for mapping the short reads to a reference genome has subsequently increased thus creating significant delays post-sequencing. To help alleviate this bottleneck TimeLogic has developed a novel short-read mapping algorithm and accelerated hardware system named VelociMapper, which can map a typical data set in a few minutes instead of hours. In order to show that our new algorithm generates results similar to the widely used alignment algorithms, BWA and Bowtie2, we used a ChIP-Seq 36-nt data set to compare the sensitivity and specificity of all three alignment algorithms. With regard to total alignment statistics, all three algorithms gave similar results (97.5%-97.9% mapping rate; 83.6%-83.8% unique aligns). Each data set contained 0.05% reads that were not mapped in the other programs and these differences could be explained by the fact that it was not possible to set perfectly identical parameters on how indels were treated. To further compare the alignment output obtained with the 3 algorithms, we ran the BAM files through Active Motif’s standard ChIP-Seq analysis pipeline. The MACS peak calling algorithm identified 20,248, 20,247, and 20,218 peaks for the VelociMapper, BWA, and Bowtie2 data sets, respectively. Correlation analysis of peak calling gave Pearson coefficients of 0.99992 between VelociMapper and BWA, and 0.99958 between VelociMapper and Bowtie2. We conclude that our dramatically accelerated algorithm gives results similar to both BWA and Bowtie2.

Short Abstract: Genes are transcriptionally regulated by complex combinations of transcription factors (TFs) in a cell-type and locus specific manner. Understanding which combinations of TFs regulate which genes has remained elusive. We integrate ENCODE ChIP-seq datasets for ~100 TFs in five cell-lines with expression data to reveal the context-specific regulatory code. To account for variability in ChIP antibody efficiency and data quality, we boost the binding signal using complementary sequence motif information and hypersensitivity data. We treat the binding profiles of TFs as a continuum of binding affinities rather than discrete binding events, applying rank-based signal transformations. Using a gene-centric view, we learn statistically significant hierarchical and disjoint ‘biclusters’ of specific TF combinations that localize to promoter regions. The complementary TF-centric view incorporates distal regions based on binding sites of all TFs. We identify biclusters that show consistent binding and those that vary across cell-lines and relate the differential co-binding with differential expression of the target genes. We find that biclusters involving a focus TF with different binding partners display functional modularity. We also find that some TFs switch their preferred partners across different cell-line and gene-proximal and distal neighborhoods. We are thus able to identify several known co-associations as well some novel surprising interactions. We also build targeted, supervised predictive models of gene expression using sequence-motifs, the TF binding profiles and TF knockdown data to corroborate the biclustering results and understand the extent to which we can explain differential gene-expression with the limited yet diverse set of high-resolution TF binding data.

Short Abstract: The NF-kappaB family plays a prominent role in the innate immune response, cell cycle activation or cell apoptosis. NF-kappaB-dependent genes can be categorized, based on the timing of their activation counted from NF-kappaB translocation into the nucleus, as Early, Middle and Late genes. It is not obvious what mechanism is responsible for segregation of the genes’ timing of transcriptional response. It is likely that the differences in timing are reflected in differences in the structure of promoter regions of genes in different categories. Specifically, this might concern differences in number and type of transcription factor binding motifs, required for NF-kappaB itself as well as for the putative cofactors. Another control sequences covered in this research are AU - rich elements (ARE) located in 3’UTR. Recent studies show that genes transcribed with unstable mRNA have different transcription dynamic. Our data suggests that the rapid response of the NF-kappaB dependent Early genes may be due to both increased gene transcription due to NF-kappaB loading as well as the contribution of mRNA instability to the transcript profiles. Wider phylogenetic analysis of NF-kappaB dependent genes provides insight into the degree of cross-species similarity found in the Early genes, opposed to many differences in promoter structure that can be found among the Late genes. This data suggest that activation and expression of the Late genes is much more species-specific than of the Early genes.

Short Abstract:

Short Abstract: The ChIP-seq has become a powerful and efficient methods to study protein-DNA interactions during gene transcriptional regulation. However, it has been a major challenge to identify the cis-regulatory elements in a ChIP-seq dataset that typically contains thousands of sequences as many popular motif discovery methods do not scale well to such large sequences datasets and lack ability to find multiple co-factor motifs. In this work, we developed an accurate and efficient discriminative motif discovery method to address these issues. In our approach, a k-mer enumeration based technique is firstly used to enrich over-presented motifs, thereby reduce the space of potential motifs, and then position weight matrices (PWMs) are constructed and updated by using a modified Gibbs sampling strategy. When tested on large simulated and real biological benchmark datasets, our method is able to very quickly detecting binding motifs of various lengths for the ChIP-ed TF and co-factors simultaneously, and it outperforms some commonly used motif discovery tools, e.g. DREME in both accuracy and efficiency.