Network inference is used to model transcriptional, signaling, and metabolic interactions among genes, proteins, and metabolites that identify biological pathways influencing disease pathogenesis. Advances in machine learning (ML)-based inference models exhibit the predictive capabilities of capturing latent patterns in genomic data. Such models are emerging as an alternative to the statistical models identifying causative factors driving complex diseases. We present CoVar, an ML-based framework that builds upon the properties of existing inference models, to find the central genes driving perturbed gene expression across biological states. Unlike differentially expressed genes (DEGs) that capture changes in individual gene expression across conditions, CoVar focuses on identifying variational genes that undergo changes in their expression network interaction profiles, providing insights into changes in the regulatory dynamics, such as in disease pathogenesis. Subsequently, it finds core genes from among the nearest neighbors of these variational genes, which are central to the variational activity and influence the coordinated regulatory processes underlying the observed changes in gene expression. Through the analysis of simulated as well as yeast expression data perturbed by the deletion of the mitochondrial genome, we show that CoVar captures the intrinsic variationality and modularity in the expression data, identifying key driver genes not found through existing differential analysis methodologies.

MOTIVATION: Analysis of open chromatin regions across multiple samples from two or more distinct conditions can determine altered gene regulatory patterns associated with biological phenotypes and complex traits. The ATAC-seq assay allows for tractable genome-wide open chromatin profiling of large numbers of samples. Stable, broadly applicable genomic annotations of open chromatin regions are not available. Thus, most studies first identify open regions using peak calling methods for each sample independently. These are then heuristically combined to obtain a consensus peak set. Reconciling sample-specific peak results post hoc from larger cohorts is particularly challenging, and informative spatial features specific to open chromatin signals are not leveraged effectively. RESULTS: We propose a novel method, ROCCO, that determines consensus open chromatin regions across multiple samples simultaneously. ROCCO employs robust summary statistics and solves a constrained optimization problem formulated to account for both enrichment and spatial dependence of open chromatin signal data. We show this formulation admits attractive theoretical and conceptual properties as well as superior empirical performance compared to current methodology. AVAILABILITY AND IMPLEMENTATION: Source code, documentation, and usage demos for ROCCO are available on GitHub at: https://github.com/nolan-h-hamilton/ROCCO. ROCCO can also be installed as a stand-alone binary utility using pip/PyPI.

Motivation: Identifying the locations of transcription factor binding sites is critical for understanding how gene transcription is regulated across different cell types and conditions. Chromatin accessibility experiments such as DNaseI sequencing (DNase-seq) and Assay for Transposase Accessible Chromatin sequencing (ATAC-seq) produce genome-wide data that include distinct 'footprint' patterns at binding sites. Nearly all existing computational methods to detect footprints from these data assume that footprint signals are highly homogeneous across footprint sites. Additionally, a comprehensive and systematic comparison of footprinting methods for specifically identifying which motif sites for a specific factor are bound has not been performed. Results: Using DNase-seq data from the ENCODE project, we show that a large degree of previously uncharacterized site-to-site variability exists in footprint signal across motif sites for a transcription factor. To model this heterogeneity in the data, we introduce a novel, supervised learning footprinter called Detecting Footprints Containing Motifs (DeFCoM). We compare DeFCoM to nine existing methods using evaluation sets from four human cell-lines and eighteen transcription factors and show that DeFCoM outperforms current methods in determining bound and unbound motif sites. We also analyze the impact of several biological and technical factors on the quality of footprint predictions to highlight important considerations when conducting footprint analyses and assessing the performance of footprint prediction methods. Finally, we show that DeFCoM can detect footprints using ATAC-seq data with similar accuracy as when using DNase-seq data. Availability and Implementation: Python code available at https://bitbucket.org/bryancquach/defcom. Contact: [email protected] or [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.

BACKGROUND: High-throughput sequence (HTS) data exhibit position-specific nucleotide biases that obscure the intended signal and reduce the effectiveness of these data for downstream analyses. These biases are particularly evident in HTS assays for identifying regulatory regions in DNA (DNase-seq, ChIP-seq, FAIRE-seq, ATAC-seq). Biases may result from many experiment-specific factors, including selectivity of DNA restriction enzymes and fragmentation method, as well as sequencing technology-specific factors, such as choice of adapters/primers and sample amplification methods. RESULTS: We present a novel method to detect and correct position-specific nucleotide biases in HTS short read data. Our method calculates read-specific weights based on aligned reads to correct the over- or underrepresentation of position-specific nucleotide subsequences, both within and adjacent to the aligned read, relative to a baseline calculated in assay-specific enriched regions. Using HTS data from a variety of ChIP-seq, DNase-seq, FAIRE-seq, and ATAC-seq experiments, we show that our weight-adjusted reads reduce the position-specific nucleotide imbalance across reads and improve the utility of these data for downstream analyses, including identification and characterization of open chromatin peaks and transcription-factor binding sites. CONCLUSIONS: A general-purpose method to characterize and correct position-specific nucleotide sequence biases fills the need to recognize and deal with, in a systematic manner, binding-site preference for the growing number of HTS-based epigenetic assays. As the breadth and impact of these biases are better understood, the availability of a standard toolkit to correct them will be important.

MOTIVATION: Advances in chromosome conformation capture and next-generation sequencing technologies are enabling genome-wide investigation of dynamic chromatin interactions. For example, Hi-C experiments generate genome-wide contact frequencies between pairs of loci by sequencing DNA segments ligated from loci in close spatial proximity. One essential task in such studies is peak calling, that is, detecting non-random interactions between loci from the two-dimensional contact frequency matrix. Successful fulfillment of this task has many important implications including identifying long-range interactions that assist interpreting a sizable fraction of the results from genome-wide association studies. The task - distinguishing biologically meaningful chromatin interactions from massive numbers of random interactions - poses great challenges both statistically and computationally. Model-based methods to address this challenge are still lacking. In particular, no statistical model exists that takes the underlying dependency structure into consideration. RESULTS: In this paper, we propose a hidden Markov random field (HMRF) based Bayesian method to rigorously model interaction probabilities in the two-dimensional space based on the contact frequency matrix. By borrowing information from neighboring loci pairs, our method demonstrates superior reproducibility and statistical power in both simulation studies and real data analysis. AVAILABILITY AND IMPLEMENTATION: The Source codes can be downloaded at: http://www.unc.edu/∼yunmli/HMRFBayesHiC CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Several bioinformatic approaches have previously been used to find novel sites of ADAR mediated A-to-I RNA editing in human. These studies have discovered thousands of genes that are hyper-edited in their non-coding intronic regions, especially in alu retrotransposable elements, but very few substrates that are site-selectively edited in coding regions. Known RNA edited substrates suggest, however, that site selective A-to-I editing is particularly important for normal brain development in mammals. RESULTS: We have compiled a screen that enables the identification of new sites of site-selective editing, primarily in coding sequences. To avoid hyper-edited repeat regions, we applied our screen to the alu-free mouse genome. Focusing on the mouse also facilitated better experimental verification. To identify candidate sites of RNA editing, we first performed an explorative screen based on RNA structure and genomic sequence conservation. We further evaluated the results of the explorative screen by determining which transcripts were enriched for A-G mismatches between the genomic template and the expressed sequence since the editing product, inosine (I), is read as guanosine (G) by the translational machinery. For expressed sequences, we only considered coding regions to focus entirely on re-coding events. Lastly, we refined the results from the explorative screen using a novel scoring scheme based on characteristics for known A-to-I edited sites. The extent of editing in the final candidate genes was verified using total RNA from mouse brain and 454 sequencing. CONCLUSIONS: Using this method, we identified and confirmed efficient editing at one site in the Gabra3 gene. Editing was also verified at several other novel sites within candidates predicted to be edited. Five of these sites are situated in genes coding for the neuron-specific RNA binding proteins HuB and HuD.

UNLABELLED: Tag sequencing using high-throughput sequencing technologies are now regularly employed to identify specific sequence features, such as transcription factor binding sites (ChIP-seq) or regions of open chromatin (DNase-seq). To intuitively summarize and display individual sequence data as an accurate and interpretable signal, we developed F-Seq, a software package that generates a continuous tag sequence density estimation allowing identification of biologically meaningful sites whose output can be displayed directly in the UCSC Genome Browser. AVAILABILITY: The software is written in the Java language and is available on all major computing platforms for download at http://www.genome.duke.edu/labs/furey/software/fseq.

MOTIVATION: Genes silenced by the aberrent methylation of nearby CpG islands can contribute to the onset or progression of cancer and represent potential biomarkers for diagnosis and prognosis. Relatively few have thus far been validated as hypermethylated in cancer among over 14,000 candidates with promoter region CpG islands. A descriptive set of genes known to be unmethylated in cancer does not exist. This lack of a negative set and a large number of candidates necessitated the development of a new approach to identify novel genes hypermethylated in cancer. RESULTS: We developed a general method, cluster_boost, that in an imbalanced data setting predicts new minority class members given limited known samples and a large set of unlabeled samples. Synthetic datasets modeled after the hypermethylated genes data show that cluster_boost can successfully identify minority samples within unlabeled data. Using genome sequence features, cluster_boost predicted candidate hypermethylated genes among 14,000 genes of unknown status. In primary ovarian cancers, we determined the methylation status for 15 genes with different levels of support for being hypermethlyated. Results indicate cluster_boost can accurately identify novel genes hypermethylated in cancer. AVAILABILITY: Software and datasets are freely available at http://labs.genome.duke.edu/FureyLab/cluster_boost.php. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

A significant number of human X-linked genes escape X chromosome inactivation and are thus expressed from both the active and inactive X chromosomes. The basis for escape from inactivation and the potential role of the X chromosome primary DNA sequence in determining a gene's X inactivation status is unclear. Using a combination of the X chromosome sequence and a comprehensive X inactivation profile of more than 600 genes, two independent yet complementary approaches were used to systematically investigate the relationship between X inactivation and DNA sequence features. First, statistical analyses revealed that a number of repeat features, including long interspersed nuclear element (LINE) and mammalian-wide interspersed repeat repetitive elements, are significantly enriched in regions surrounding transcription start sites of genes that are subject to inactivation, while Alu repetitive elements and short motifs containing ACG/CGT are significantly enriched in those that escape inactivation. Second, linear support vector machine classifiers constructed using primary DNA sequence features were used to correctly predict the X inactivation status for >80% of all X-linked genes. We further identified a small set of features that are important for accurate classification, among which LINE-1 and LINE-2 content show the greatest individual discriminatory power. Finally, as few as 12 features can be used for accurate support vector machine classification. Taken together, these results suggest that features of the underlying primary DNA sequence of the human X chromosome may influence the spreading and/or maintenance of X inactivation.

MOTIVATION: DNA microarray experiments generating thousands of gene expression measurements, are being used to gather information from tissue and cell samples regarding gene expression differences that will be useful in diagnosing disease. We have developed a new method to analyse this kind of data using support vector machines (SVMs). This analysis consists of both classification of the tissue samples, and an exploration of the data for mis-labeled or questionable tissue results. RESULTS: We demonstrate the method in detail on samples consisting of ovarian cancer tissues, normal ovarian tissues, and other normal tissues. The dataset consists of expression experiment results for 97,802 cDNAs for each tissue. As a result of computational analysis, a tissue sample is discovered and confirmed to be wrongly labeled. Upon correction of this mistake and the removal of an outlier, perfect classification of tissues is achieved, but not with high confidence. We identify and analyse a subset of genes from the ovarian dataset whose expression is highly differentiated between the types of tissues. To show robustness of the SVM method, two previously published datasets from other types of tissues or cells are analysed. The results are comparable to those previously obtained. We show that other machine learning methods also perform comparably to the SVM on many of those datasets. AVAILABILITY: The SVM software is available at http://www.cs. columbia.edu/ approximately bgrundy/svm.