Colocalization analyses of genomic elements: approaches, recommendations and challenges

The causal role of circulating inflammatory markers in osteoporosis: a bidirectional Mendelian randomized study

Background

Osteoporosis (OP) associated with aging exerts substantial clinical and fiscal strains on societal structures. An increasing number of research studies have suggested a bidirectional relationship between circulating inflammatory markers (CIMs) and OP. However, observational studies are susceptible to perturbations in confounding variables. In contrast, Mendelian randomization (MR) offers a robust methodological framework to circumvent such confounders, facilitating a more accurate assessment of causality. Our study aimed to evaluate the causal relationships between CIMs and OP, identifying new approaches and strategies for the prevention, diagnosis and treatment of OP.

Methods

We analyzed publicly available GWAS summary statistics to investigate the causal relationships between CIMs and OP. Causal estimates were calculated via a systematic analytical framework, including bidirectional MR analysis and Bayesian colocalization analysis.

Results

Genetically determined levels of CXCL11 (OR = 0.91, 95% CI = 0.85-0.98, P = 0.008, PFDR = 0.119), IL-18 (OR = 0.88, 95% CI = 0.83-0.94, P = 8.66×10-5, PFDR = 0.008), and LIF (OR = 0.86, 95% CI = 0.76-0.96, P = 0.008, PFDR = 0.119) were linked to a reduced risk of OP. Conversely, higher levels of ARTN (OR = 1.11, 95% CI = 1.02-1.20, P = 0.012, PFDR = 0.119) and IFNG (OR = 1.16, 95% CI = 1.03-1.30, P = 0.013, PFDR = 0.119) were associated with an increased risk of OP. Bayesian colocalization analysis revealed no evidence of shared causal variants.

Conclusion

Despite finding no overall association between CIMs and OP, five CIMs demonstrated a potentially significant association with OP. These findings could pave the way for future mechanistic studies aimed at discovering new treatments for this disease. Additionally, we are the first to suggest a unidirectional causal relationship between ARTN and OP. This novel insight introduces new avenues for research into diagnostic and therapeutic strategies for OP.

Global and local genomic features together modulate the spontaneous single nucleotide mutation rate

Spontaneous mutations are evolutionary engines as they generate variants for the evolutionary downstream processes that give rise to speciation and adaptation. Single nucleotide mutations (SNM) are the most abundant type of mutations among them. Here, we perform a meta-analysis to quantify the influence of selected global genomic parameters (genome size, genomic GC content, genomic repeat fraction, number of coding genes, gene count, and strand bias in prokaryotes) and local genomic features (local GC content, repeat content, CpG content and the number of SNM at CpG islands) on spontaneous SNM rates across the tree of life (prokaryotes, unicellular eukaryotes, multicellular eukaryotes) using wild-type sequence data in two different taxon classification systems. We find that the spontaneous SNM rates in our data are correlated with many genomic features in prokaryotes and unicellular eukaryotes irrespective of their sample sizes. On the other hand, only the number of coding genes was correlated with the spontaneous SNM rates in multicellular eukaryotes primarily contributed by vertebrates data. Considering local features, we notice that local GC content and CpG content significantly were correlated with the spontaneous SNM rates in the unicellular eukaryotes, while local repeat fraction is an important feature in prokaryotes and certain specific uni- and multi-cellular eukaryotes. Such predictive features of the spontaneous SNM rates often support non-linear models as the best fit compared to the linear model. We also observe that the strand asymmetry in prokaryotes plays an important role in determining the spontaneous SNM rates but the SNM spectrum does not.

Efficient Analysis of Annotation Colocalization Accounting for Genomic Contexts

An annotation is a set of genomic intervals sharing a particular function or property. Examples include genes or their exons, evolutionarily conserved elements, and regions with a particular epigenetic state. A common task is to compare two annotations to determine if one is enriched or depleted in the regions covered by the other. We study the problem of assigning statistical significance to such a comparison based on a null model representing two random unrelated annotations. To incorporate more background information into such analyses,we propose a new null model based on a Markov chain which differentiates among several genomic contexts. These contexts can capture various confounding factors, such as GC content or assembly gaps. We then develop a new algorithm for estimating p-values by computing the exact expectation and variance of the test statistics and then estimating the p-value using a normal approximation. Compared to the previous algorithm by Gafurov et al., the new algorithm provides three advances: (1) the running time is improved from quadratic to linear or quasi-linear, (2) the algorithm can handle two different test statistics, and (3) the algorithm can handle both simple and context-dependent Markov chain null models. We demonstrate the efficiency and accuracy of our algorithm on synthetic and real data sets, including the recent human telomere-to-telomere assembly. In particular, our algorithm computed p-values for 450 pairs of human genome annotations using 24 threads in under three hours. Moreover, the use of genomic contexts to correct for GC bias resulted in the reversal of some previously published findings.

Association between rheumatoid arthritis and autoimmune thyroid disease: evidence from complementary genetic methods

OBJECTIVES

To assess the causal association of Rheumatoid Arthritis (RA) with Autoimmune thyroid disease (AITD).

METHOD

Complementary genetic approaches, including genetic correlation, Mendelian randomization (MR) and colocalization analysis, were conducted to assess the potential causal association between RA and AITD using summary statistics from large-scale genome-wide association studies (GWASs). Various sensitivity analyses had been conducted to assess the robustness and the consistency of the findings.

RESULTS

The linkage disequilibrium score regression revealed a shared genetic structure between RA and AITD, with the significant genetic correlation between RA and autoimmune hyperthyroidism and autoimmune hypothyroidism estimated to be 0.3945 (P = 2.83 × 10-6) and 0.2771 (P = 1.04 × 10-6) respectively. The results of MR analysis showed that RA had a positive causal relationship with autoimmune hypothyroidism and autoimmune hyperthyroidism. The odds ratio (OR) were 1.29 (95% CI, 1.17-1.42; P = 1.08 × 10-7) and 1.47 (95% CI, 1.25-1.72; P = 1.85 × 10-6), respectively. In reverse MR analysis, autoimmune hypothyroidism had a positive causal relationship with RA, OR was 1.51 (95% CI, 1.37-1.66; P = 1.10 × 10-16); autoimmune hyperthyroidism had no causal relationship with RA relationship (P = 0.22). Similar results were found using different MR methods. In addition, colocalization analysis suggested that shared causal variants existed between RA and AITD.

CONCLUSIONS

Our study suggested a potentially causal effect of genetically predicted RA on autoimmune hyperthyroidism and a bidirectional causal relationship between RA and autoimmune hypothyroidism was also observed with complementary genetic approaches, which supports the importance and necessity of thyroid function screening and monitoring in RA patient management in clinical practice.

MoCoLo: a testing framework for motif co-localization

Sequence-level data offers insights into biological processes through the interaction of two or more genomic features from the same or different molecular data types. Within motifs, this interaction is often explored via the co-occurrence of feature genomic tracks using fixed-segments or analytical tests that respectively require window size determination and risk of false positives from over-simplified models. Moreover, methods for robustly examining the co-localization of genomic features, and thereby understanding their spatial interaction, have been elusive. We present a new analytical method for examining feature interaction by introducing the notion of reciprocal co-occurrence, define statistics to estimate it and hypotheses to test for it. Our approach leverages conditional motif co-occurrence events between features to infer their co-localization. Using reverse conditional probabilities and introducing a novel simulation approach that retains motif properties (e.g. length, guanine-content), our method further accounts for potential confounders in testing. As a proof-of-concept, motif co-localization (MoCoLo) confirmed the co-occurrence of histone markers in a breast cancer cell line. As a novel analysis, MoCoLo identified significant co-localization of oxidative DNA damage within non-B DNA-forming regions that significantly differed between non-B DNA structures. Altogether, these findings demonstrate the potential utility of MoCoLo for testing spatial interactions between genomic features via their co-localization.

Integrated analysis of Mendelian Randomization and Bayesian colocalization reveals bidirectional causal association between inflammatory bowel disease and psoriasis

BACKGROUND

Observational studies have suggested an association between inflammatory bowel disease [IBD] and psoriasis. However, the detailed genetic basis, causality, and direction of this association remain unclear.

METHODS

Bidirectional two-sample Mendelian Randomization [MR] analysis was conducted using summary statistics from published genome-wide association studies. Bayesian Colocalization and multivariable MR [MVMR] analyses were performed to identify candidate variants and risk genes involved in the shared genetic basis between IBD, psoriasis, and their subtypes.

RESULTS

Genetically predicted IBD and Crohn's disease [CD] were associated with an increased risk of psoriasis, psoriasis vulgaris [PsV], and psoriatic arthritis [PsA] (IBD on psoriasis: pooled odds ratio [OR] 1.09, 95% confidence interval [CI] 1.04-1.14, p = .0001; CD on psoriasis: pooled OR 1.10, 95% CI 1.06-1.15, p < .0001) and vice versa (psoriasis on IBD: pooled OR 1.11, 95%CI 1.02-1.21), whereas CD only exhibited a unidirectional association with psoriasis. Colocalization analysis revealed eight candidate genetic variants and risk genes (including LINC00824, CDKAL1, IL10, IL23R, DNAJC27, LPP, RUNX3, and RGS14) associated with a shared genetic basis. Among these, IL23R, DNAJC27, LPP, and RGS14 were further validated by MVMR analysis.

CONCLUSION

Our findings indicated bidirectional causal associations between IBD and psoriasis (including PsV and PsA), which were attributed primarily to CD rather than Ulcerative colitis [UC]. Furthermore, we identified several candidate variants and risk genes involved in the shared genetic basis of IBD and psoriasis. Acquiring a better understanding of the shared genetic architecture underlying IBD and psoriasis would help improve clinical strategies.

Appraising the Causal Role of Risk Factors in Coronary Artery Disease and Stroke: A Systematic Review of Mendelian Randomization Studies

BACKGROUND Mendelian randomization (MR) offers a powerful approach to study potential causal associations between exposures and health outcomes by using genetic variants associated with an exposure as instrumental variables. In this systematic review, we aimed to summarize previous MR studies and to evaluate the evidence for causality for a broad range of exposures in relation to coronary artery disease and stroke. METHODS AND RESULTS MR studies investigating the association of any genetically predicted exposure with coronary artery disease or stroke were identified. Studies were classified into 4 categories built on the significance of the main MR analysis results and its concordance with sensitivity analyses, namely, robust, probable, suggestive, and insufficient. Studies reporting associations that did not perform any sensitivity analysis were classified as nonevaluable. We identified 2725 associations eligible for evaluation, examining 535 distinct exposures. Of them, 141 were classified as robust, 353 as probable, 110 as suggestive, and 926 had insufficient evidence. The most robust associations were observed for anthropometric traits, lipids, and lipoproteins and type 2 diabetes with coronary artery; disease and clinical measurements with coronary artery disease and stroke; and thrombotic factors with stroke. CONCLUSIONS Despite the large number of studies that have been conducted, only a limited number of associations were supported by robust evidence. Approximately half of the studies reporting associations presented an MR sensitivity analysis along with the main analysis that further supported the causality of associations. Future research should focus on more thorough assessments of sensitivity MR analyses and further assessments of mediation effects or nonlinearity of associations.

RgnTX: Colocalization analysis of transcriptome elements in the presence of isoform heterogeneity and ambiguity

Colocalization analysis of genomic region sets has been widely adopted to unveil potential functional interactions between corresponding biological attributes, which often serves as the basis for further investigation. A number of methods have been developed for colocalization analysis of genomic elements. However, none of them explicitly considered the transcriptome heterogeneity and isoform ambiguity, making them less appropriate for analyzing transcriptome elements. Here, we developed RgnTX, an R/Bioconductor tool for the colocalization analysis of transcriptome elements with permutation tests. Different from existing approaches, RgnTX directly takes advantage of transcriptome annotation, and offers high flexibility in the null model to simulate realistic transcriptome-wide background, such as the complex alternative splicing patterns. Importantly, it supports the testing of transcriptome elements without clear isoform association, which is often the real scenario due to technical limitations. Proposed package offers a wide selection of pre-defined functions, easy to be utilized by users for visualizing permutation results, calculating shifted z-scores and conducting multiple hypothesis testing under Benjamini-Hochberg correction. Moreover, with synthetic and real datasets, we show that RgnTX novel testing modes return distinct and more significant results compared to existing genome-based methods. We believe RgnTX should make a useful tool to characterize the randomness of the transcriptome, and for conducting statistical association analysis for genomic region sets within the heterogeneous transcriptome. The package now has been accepted by Bioconductor and is freely available at: https://bioconductor.org/packages/RgnTX.

Genome-wide association studies: utility and limitations for research in physiology

Physiological systems are subject to interindividual variation encoded by genetics. Genome-wide association studies (GWAS) operate by surveying thousands of genetic variants from a substantial number of individuals and assessing their association to a trait of interest, be it a physiological variable, a molecular phenotype (e.g. gene expression), or even a disease or condition. Through a myriad of methods, GWAS downstream analyses then explore the functional consequences of each variant and attempt to ascertain a causal relationship to the phenotype of interest, as well as to delve into its links to other traits. This type of investigation allows mechanistic insights into physiological functions, pathological disturbances and shared biological processes between traits (i.e. pleiotropy). An exciting example is the discovery of a new thyroid hormone transporter (SLC17A4) and hormone metabolising enzyme (AADAT) from a GWAS on free thyroxine levels. Therefore, GWAS have substantially contributed with insights into physiology and have been shown to be useful in unveiling the genetic control underlying complex traits and pathological conditions; they will continue to do so with global collaborations and advances in genotyping technology. Finally, the increasing number of trans-ancestry GWAS and initiatives to include ancestry diversity in genomics will boost the power for discoveries, making them also applicable to non-European populations.

bootRanges: flexible generation of null sets of genomic ranges for hypothesis testing

MOTIVATION

Enrichment analysis is a widely utilized technique in genomic analysis that aims to determine if there is a statistically significant association between two sets of genomic features. To conduct this type of hypothesis testing, an appropriate null model is typically required. However, the null distribution that is commonly used can be overly simplistic and may result in inaccurate conclusions.

RESULTS

bootRanges provides fast functions for generation of block bootstrapped genomic ranges representing the null hypothesis in enrichment analysis. As part of a modular workflow, bootRanges offers greater flexibility for computing various test statistics leveraging other Bioconductor packages. We show that shuffling or permutation schemes may result in overly narrow test statistic null distributions and over-estimation of statistical significance, while creating new range sets with a block bootstrap preserves local genomic correlation structure and generates more reliable null distributions. It can also be used in more complex analyses, such as accessing correlations between cis-regulatory elements (CREs) and genes across cell types or providing optimized thresholds, e.g. log fold change (logFC) from differential analysis.

AVAILABILITY AND IMPLEMENTATION

bootRanges is freely available in the R/Bioconductor package nullranges hosted at https://bioconductor.org/packages/nullranges.

A comparison of the genes and genesets identified by GWAS and EWAS of fifteen complex traits

Identifying genomic regions pertinent to complex traits is a common goal of genome-wide and epigenome-wide association studies (GWAS and EWAS). GWAS identify causal genetic variants, directly or via linkage disequilibrium, and EWAS identify variation in DNA methylation associated with a trait. While GWAS in principle will only detect variants due to causal genes, EWAS can also identify genes via confounding, or reverse causation. We systematically compare GWAS (N > 50,000) and EWAS (N > 4500) results of 15 complex traits. We evaluate if the genes or gene ontology terms flagged by GWAS and EWAS overlap, and find substantial overlap for diastolic blood pressure, (gene overlap P = 5.2 × 10^-6; term overlap P = 0.001). We superimpose our empirical findings against simulated models of varying genetic and epigenetic architectures and observe that in most cases GWAS and EWAS are likely capturing distinct genesets. Our results indicate that GWAS and EWAS are capturing different aspects of the biology of complex traits.

Genome-wide association and multi-trait analyses characterize the common genetic architecture of heart failure

Heart failure is a leading cause of cardiovascular morbidity and mortality. However, the contribution of common genetic variation to heart failure risk has not been fully elucidated, particularly in comparison to other common cardiometabolic traits. We report a multi-ancestry genome-wide association study meta-analysis of all-cause heart failure including up to 115,150 cases and 1,550,331 controls of diverse genetic ancestry, identifying 47 risk loci. We also perform multivariate genome-wide association studies that integrate heart failure with related cardiac magnetic resonance imaging endophenotypes, identifying 61 risk loci. Gene-prioritization analyses including colocalization and transcriptome-wide association studies identify known and previously unreported candidate cardiomyopathy genes and cellular processes, which we validate in gene-expression profiling of failing and healthy human hearts. Colocalization, gene expression profiling, and Mendelian randomization provide convergent evidence for the roles of BCKDHA and circulating branch-chain amino acids in heart failure and cardiac structure. Finally, proteome-wide Mendelian randomization identifies 9 circulating proteins associated with heart failure or quantitative imaging traits. These analyses highlight similarities and differences among heart failure and associated cardiovascular imaging endophenotypes, implicate common genetic variation in the pathogenesis of heart failure, and identify circulating proteins that may represent cardiomyopathy treatment targets.

Systematic review of Mendelian randomization studies on risk of cancer

BACKGROUND

We aimed to map and describe the current state of Mendelian randomization (MR) literature on cancer risk and to identify associations supported by robust evidence.

METHODS

We searched PubMed and Scopus up to 06/10/2020 for MR studies investigating the association of any genetically predicted risk factor with cancer risk. We categorized the reported associations based on a priori designed levels of evidence supporting a causal association into four categories, namely robust, probable, suggestive, and insufficient, based on the significance and concordance of the main MR analysis results and at least one of the MR-Egger, weighed median, MRPRESSO, and multivariable MR analyses. Associations not presenting any of the aforementioned sensitivity analyses were not graded.

RESULTS

We included 190 publications reporting on 4667 MR analyses. Most analyses (3200; 68.6%) were not accompanied by any of the assessed sensitivity analyses. Of the 1467 evaluable analyses, 87 (5.9%) were supported by robust, 275 (18.7%) by probable, and 89 (6.1%) by suggestive evidence. The most prominent robust associations were observed for anthropometric indices with risk of breast, kidney, and endometrial cancers; circulating telomere length with risk of kidney, lung, osteosarcoma, skin, thyroid, and hematological cancers; sex steroid hormones and risk of breast and endometrial cancer; and lipids with risk of breast, endometrial, and ovarian cancer.

CONCLUSIONS

Despite the large amount of research on genetically predicted risk factors for cancer risk, limited associations are supported by robust evidence for causality. Most associations did not present a MR sensitivity analysis and were thus non-evaluable. Future research should focus on more thorough assessment of sensitivity MR analyses and on more transparent reporting.

Circulating inflammatory cytokines and risk of five cancers: a Mendelian randomization analysis

BACKGROUND

Epidemiological and experimental evidence has linked chronic inflammation to cancer aetiology. It is unclear whether associations for specific inflammatory biomarkers are causal or due to bias. In order to examine whether altered genetically predicted concentration of circulating cytokines are associated with cancer development, we performed a two-sample Mendelian randomisation (MR) analysis.

METHODS

Up to 31,112 individuals of European descent were included in genome-wide association study (GWAS) meta-analyses of 47 circulating cytokines. Single nucleotide polymorphisms (SNPs) robustly associated with the cytokines, located in or close to their coding gene (cis), were used as instrumental variables. Inverse-variance weighted MR was used as the primary analysis, and the MR assumptions were evaluated in sensitivity and colocalization analyses and a false discovery rate (FDR) correction for multiple comparisons was applied. Corresponding germline GWAS summary data for five cancer outcomes (breast, endometrial, lung, ovarian, and prostate), and their subtypes were selected from the largest cancer-specific GWASs available (cases ranging from 12,906 for endometrial to 133,384 for breast cancer).

RESULTS

There was evidence of inverse associations of macrophage migration inhibitory factor with breast cancer (OR per SD = 0.88, 95% CI 0.83 to 0.94), interleukin-1 receptor antagonist with endometrial cancer (0.86, 0.80 to 0.93), interleukin-18 with lung cancer (0.87, 0.81 to 0.93), and beta-chemokine-RANTES with ovarian cancer (0.70, 0.57 to 0.85) and positive associations of monokine induced by gamma interferon with endometrial cancer (3.73, 1.86 to 7.47) and cutaneous T-cell attracting chemokine with lung cancer (1.51, 1.22 to 1.87). These associations were similar in sensitivity analyses and supported in colocalization analyses.

CONCLUSIONS

Our study adds to current knowledge on the role of specific inflammatory biomarker pathways in cancer aetiology. Further validation is needed to assess the potential of these cytokines as pharmacological or lifestyle targets for cancer prevention.

OLOGRAM-MODL: mining enriched n-wise combinations of genomic features with Monte Carlo and dictionary learning

Most epigenetic marks, such as Transcriptional Regulators or histone marks, are biological objects known to work together in n-wise complexes. A suitable way to infer such functional associations between them is to study the overlaps of the corresponding genomic regions. However, the problem of the statistical significance of n-wise overlaps of genomic features is seldom tackled, which prevent rigorous studies of n-wise interactions. We introduce OLOGRAM-MODL, which considers overlaps between n ≥ 2 sets of genomic regions, and computes their statistical mutual enrichment by Monte Carlo fitting of a Negative Binomial distribution, resulting in more resolutive P-values. An optional machine learning method is proposed to find complexes of interest, using a new itemset mining algorithm based on dictionary learning which is resistant to noise inherent to biological assays. The overall approach is implemented through an easy-to-use CLI interface for workflow integration, and a visual tree-based representation of the results suited for explicability. The viability of the method is experimentally studied using both artificial and biological data. This approach is accessible through the command line interface of the pygtftk toolkit, available on Bioconda and from https://github.com/dputhier/pygtftk

OUP accepted manuscript

Motivation

Genome annotations are a common way to represent genomic features such as genes, regulatory elements or epigenetic modifications. The amount of overlap between two annotations is often used to ascertain if there is an underlying biological connection between them. In order to distinguish between true biological association and overlap by pure chance, a robust measure of significance is required. One common way to do this is to determine if the number of intervals in the reference annotation that intersect the query annotation is statistically significant. However, currently employed statistical frameworks are often either inefficient or inaccurate when computing P-values on the scale of the whole human genome.

Results

We show that finding the P-values under the typically used ‘gold’ null hypothesis is NP-hard. This motivates us to reformulate the null hypothesis using Markov chains. To be able to measure the fidelity of our Markovian null hypothesis, we develop a fast direct sampling algorithm to estimate the P-value under the gold null hypothesis. We then present an open-source software tool MCDP that computes the P-values under the Markovian null hypothesis in O(m2+n) time and O(m) memory, where m and n are the numbers of intervals in the reference and query annotations, respectively. Notably, MCDP runtime and memory usage are independent from the genome length, allowing it to outperform previous approaches in runtime and memory usage by orders of magnitude on human genome annotations, while maintaining the same level of accuracy.

Availability and implementation

The software is available at https://github.com/fmfi-compbio/mc-overlaps. All data for reproducibility are available at https://github.com/fmfi-compbio/mc-overlaps-reproducibility.

Supplementary information

Supplementary data are available at Bioinformatics online.

Performing post-genome-wide association study analysis: overview, challenges and recommendations

Genome-wide association studies (GWAS) provide  huge information on statistically significant single-nucleotide polymorphisms (SNPs) associated with various human complex traits and diseases. By performing GWAS studies, scientists have successfully identified the association of hundreds of thousands to  millions of SNPs to a single phenotype. Moreover, the association of some SNPs with rare diseases has been intensively tested. However, classic GWAS studies have not yet provided solid, knowledgeable insight into functional and biological mechanisms underlying phenotypes or mechanisms of diseases. Therefore, several post-GWAS (pGWAS) methods have been recommended. Currently, there is no simple scientific document to provide a quick guide for performing pGWAS analysis. pGWAS is a crucial step for a better understanding of the biological machinery beyond the SNPs. Here, we provide an overview to performing pGWAS analysis and demonstrate the challenges behind each method. Furthermore, we direct readers to key articles for each pGWAS method and present the overall issues in pGWAS analysis.  Finally, we include a custom pGWAS pipeline to guide new users when performing their research.

Bedshift: perturbation of genomic interval sets

Functional genomics experiments, like ChIP-Seq or ATAC-Seq, produce results that are summarized as a region set. There is no way to objectively evaluate the effectiveness of region set similarity metrics. We present Bedshift, a tool for perturbing BED files by randomly shifting, adding, and dropping regions from a reference file. The perturbed files can be used to benchmark similarity metrics, as well as for other applications. We highlight differences in behavior between metrics, such as that the Jaccard score is most sensitive to added or dropped regions, while coverage score is most sensitive to shifted regions.

VariantStore: an index for large-scale genomic variant search

Efficiently scaling genomic variant search indexes to thousands of samples is computationally challenging due to the presence of multiple coordinate systems to avoid reference biases. We present VariantStore, a system that indexes genomic variants from multiple samples using a variation graph and enables variant queries across any sample-specific coordinate system. We show the scalability of VariantStore by indexing genomic variants from the TCGA project in 4 h and the 1000 Genomes project in 3 h. Querying for variants in a gene takes between 0.002 and 3 seconds using memory only 10% of the size of the full representation.

Embeddings of genomic region sets capture rich biological associations in lower dimensions

MOTIVATION

Genomic region sets summarize functional genomics data and define locations of interest in the genome such as regulatory regions or transcription factor binding sites. The number of publicly available region sets has increased dramatically, leading to challenges in data analysis.

RESULTS

We propose a new method to represent genomic region sets as vectors, or embeddings, using an adapted word2vec approach. We compared our approach to two simpler methods based on interval unions or term frequency-inverse document frequency and evaluated the methods in three ways: First, by classifying the cell line, antibody, or tissue type of the region set; second, by assessing whether similarity among embeddings can reflect simulated random perturbations of genomic regions; and third, by testing robustness of the proposed representations to different signal thresholds for calling peaks. Our word2vec-based region set embeddings reduce dimensionality from more than a hundred thousand to 100 without significant loss in classification performance. The vector representation could identify cell line, antibody, and tissue type with over 90% accuracy. We also found that the vectors could quantitatively summarize simulated random perturbations to region sets and are more robust to subsampling the data derived from different peak calling thresholds. Our evaluations demonstrate that the vectors retain useful biological information in relatively lower-dimensional spaces. We propose that vector representation of region sets is a promising approach for efficient analysis of genomic region data.

AVAILABILITY

https://github.com/databio/regionset-embedding.

Widespread Exaptation of L1 Transposons for Transcription Factor Binding in Breast Cancer

L1 transposons occupy 17% of the human genome and are widely exapted for the regulation of human genes, particularly in breast cancer, where we have previously shown abundant cancer-specific transcription factor (TF) binding sites within the L1PA2 subfamily. In the current study, we performed a comprehensive analysis of TF binding activities in primate-specific L1 subfamilies and identified pervasive exaptation events amongst these evolutionarily related L1 transposons. By motif scanning, we predicted diverse and abundant TF binding potentials within the L1 transposons. We confirmed substantial TF binding activities in the L1 subfamilies using TF binding sites consolidated from an extensive collection of publicly available ChIP-seq datasets. Young L1 subfamilies (L1HS, L1PA2 and L1PA3) contributed abundant TF binding sites in MCF7 cells, primarily via their 5' UTR. This is expected as the L1 5' UTR hosts cis-regulatory elements that are crucial for L1 replication and mobilisation. Interestingly, the ancient L1 subfamilies, where 5' truncation was common, displayed comparable TF binding capacity through their 3' ends, suggesting an alternative exaptation mechanism in L1 transposons that was previously unnoticed. Overall, primate-specific L1 transposons were extensively exapted for TF binding in MCF7 breast cancer cells and are likely prominent genetic players modulating breast cancer transcriptional regulation.

Recommendations for the FAIRification of genomic track metadata

Background: Many types of data from genomic analyses can be represented as genomic tracks, i.e. features linked to the genomic coordinates of a reference genome. Examples of such data are epigenetic DNA methylation data, ChIP-seq peaks, germline or somatic DNA variants, as well as RNA-seq expression levels. Researchers often face difficulties in locating, accessing and combining relevant tracks from external sources, as well as locating the raw data, reducing the value of the generated information.

Description of work: We propose to advance the application of FAIR data principles (Findable, Accessible, Interoperable, and Reusable) to produce searchable metadata for genomic tracks. Findability and Accessibility of metadata can then be ensured by a track search service that integrates globally identifiable metadata from various track hubs in the Track Hub Registry and other relevant repositories. Interoperability and Reusability need to be ensured by the specification and implementation of a basic set of recommendations for metadata. We have tested this concept by developing such a specification in a JSON Schema, called FAIRtracks, and have integrated it into a novel track search service, called TrackFind. We demonstrate practical usage by importing datasets through TrackFind into existing examples of relevant analytical tools for genomic tracks: EPICO and the GSuite HyperBrowser.

Conclusion: We here provide a first iteration of a draft standard for genomic track metadata, as well as the accompanying software ecosystem. It can easily be adapted or extended to future needs of the research community regarding data, methods and tools, balancing the requirements of both data submitters and analytical end-users.

IGD: high-performance search for large-scale genomic interval datasets

SUMMARY

Databases of large-scale genome projects now contain thousands of genomic interval datasets. These data are a critical resource for understanding the function of DNA. However, our ability to examine and integrate interval data of this scale is limited. Here, we introduce the integrated genome database (IGD), a method and tool for searching genome interval datasets more than three orders of magnitude faster than existing approaches, while using only one hundredth of the memory. IGD uses a novel linear binning method that allows us to scale analysis to billions of genomic regions.

AVAILABILITY

https://github.com/databio/IGD.

Ultrafast and scalable variant annotation and prioritization with big functional genomics data

The advances of large-scale genomics studies have enabled compilation of cell type–specific, genome-wide DNA functional elements at high resolution. With the growing volume of functional annotation data and sequencing variants, existing variant annotation algorithms lack the efficiency and scalability to process big genomic data, particularly when annotating whole-genome sequencing variants against a huge database with billions of genomic features. Here, we develop VarNote to rapidly annotate genome-scale variants in large and complex functional annotation resources. Equipped with a novel index system and a parallel random-sweep searching algorithm, VarNote shows substantial performance improvements (two to three orders of magnitude) over existing algorithms at different scales. It supports both region-based and allele-specific annotations and introduces advanced functions for the flexible extraction of annotations. By integrating massive base-wise and context-dependent annotations in the VarNote framework, we introduce three efficient and accurate pipelines to prioritize the causal regulatory variants for common diseases, Mendelian disorders, and cancers.

Genetic variant effects on gene expression in human pancreatic islets and their implications for T2D

Most signals detected by genome-wide association studies map to non-coding sequence and their tissue-specific effects influence transcriptional regulation. However, key tissues and cell-types required for functional inference are absent from large-scale resources. Here we explore the relationship between genetic variants influencing predisposition to type 2 diabetes (T2D) and related glycemic traits, and human pancreatic islet transcription using data from 420 donors. We find: (a) 7741 cis-eQTLs in islets with a replication rate across 44 GTEx tissues between 40% and 73%; (b) marked overlap between islet cis-eQTL signals and active regulatory sequences in islets, with reduced eQTL effect size observed in the stretch enhancers most strongly implicated in GWAS signal location; (c) enrichment of islet cis-eQTL signals with T2D risk variants identified in genome-wide association studies; and (d) colocalization between 47 islet cis-eQTLs and variants influencing T2D or glycemic traits, including DGKB and TCF7L2. Our findings illustrate the advantages of performing functional and regulatory studies in disease relevant tissues.

A single ChIP-seq dataset is sufficient for comprehensive analysis of motifs co-occurrence with MCOT package

Recognition of composite elements consisting of two transcription factor binding sites gets behind the studies of tissue-, stage- and condition-specific transcription. Genome-wide data on transcription factor binding generated with ChIP-seq method facilitate an identification of composite elements, but the existing bioinformatics tools either require ChIP-seq datasets for both partner transcription factors, or omit composite elements with motifs overlapping. Here we present an universal Motifs Co-Occurrence Tool (MCOT) that retrieves maximum information about overrepresented composite elements from a single ChIP-seq dataset. This includes homo- and heterotypic composite elements of four mutual orientations of motifs, separated with a spacer or overlapping, even if recognition of motifs within composite element requires various stringencies. Analysis of 52 ChIP-seq datasets for 18 human transcription factors confirmed that for over 60% of analyzed datasets and transcription factors predicted co-occurrence of motifs implied experimentally proven protein-protein interaction of respecting transcription factors. Analysis of 164 ChIP-seq datasets for 57 mammalian transcription factors showed that abundance of predicted composite elements with an overlap of motifs compared to those with a spacer more than doubled; and they had 1.5-fold increase of asymmetrical pairs of motifs with one more conservative 'leading' motif and another one 'guided'.

PheGWAS: a new dimension to visualize GWAS across multiple phenotypes

MOTIVATION

PheGWAS was developed to enhance exploration of phenome-wide pleiotropy at the genome-wide level through the efficient generation of a dynamic visualization combining Manhattan plots from GWAS with PheWAS to create a 3D 'landscape'. Pleiotropy in sub-surface GWAS significance strata can be explored in a sectional view plotted within user defined levels. Further complexity reduction is achieved by confining to a single chromosomal section. Comprehensive genomic and phenomic coordinates can be displayed.

RESULTS

PheGWAS is demonstrated using summary data from Global Lipids Genetics Consortium GWAS across multiple lipid traits. For single and multiple traits PheGWAS highlighted all 88 and 69 loci, respectively. Further, the genes and SNPs reported in Global Lipids Genetics Consortium were identified using additional functions implemented within PheGWAS. Not only is PheGWAS capable of identifying independent signals but also provides insights to local genetic correlation (verified using HESS) and in identifying the potential regions that share causal variants across phenotypes (verified using colocalization tests).

AVAILABILITY AND IMPLEMENTATION

The PheGWAS software and code are freely available at (https://github.com/georgeg0/PheGWAS).

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Reaching the End-Game for GWAS: Machine Learning Approaches for the Prioritization of Complex Disease Loci

Genome-wide association studies (GWAS) have revealed thousands of genetic loci that underpin the complex biology of many human traits. However, the strength of GWAS - the ability to detect genetic association by linkage disequilibrium (LD) - is also its limitation. Whilst the ever-increasing study size and improved design have augmented the power of GWAS to detect effects, differentiation of causal variants or genes from other highly correlated genes associated by LD remains the real challenge. This has severely hindered the biological insights and clinical translation of GWAS findings. Although thousands of disease susceptibility loci have been reported, causal genes at these loci remain elusive. Machine learning (ML) techniques offer an opportunity to dissect the heterogeneity of variant and gene signals in the post-GWAS analysis phase. ML models for GWAS prioritization vary greatly in their complexity, ranging from relatively simple logistic regression approaches to more complex ensemble models such as random forests and gradient boosting, as well as deep learning models, i.e., neural networks. Paired with functional validation, these methods show important promise for clinical translation, providing a strong evidence-based approach to direct post-GWAS research. However, as ML approaches continue to evolve to meet the challenge of causal gene identification, a critical assessment of the underlying methodologies and their applicability to the GWAS prioritization problem is needed. This review investigates the landscape of ML applications in three parts: selected models, input features, and output model performance, with a focus on prioritizations of complex disease associated loci. Overall, we explore the contributions ML has made towards reaching the GWAS end-game with consequent wide-ranging translational impact.

Co-localization between Sequence Constraint and Epigenomic Information Improves Interpretation of Whole-Genome Sequencing Data

The identification of functional regions in the noncoding human genome is difficult but critical in order to gain understanding of the role noncoding variation plays in gene regulation in human health and disease. We describe here a co-localization approach that aims to identify constrained sequences that co-localize with tissue- or cell-type-specific regulatory regions, and we show that the resulting score is particularly well suited for the identification of rare regulatory variants. For 127 tissues and cell types in the ENCODE/Roadmap Epigenomics Project, we provide catalogs of putative tissue- or cell-type-specific regulatory regions under sequence constraint. We use the newly developed co-localization score for brain tissues to score de novo mutations in whole genomes from 1,902 individuals affected with autism spectrum disorder (ASD) and their unaffected siblings in the Simons Simplex Collection. We show that noncoding de novo mutations near genes co-expressed in midfetal brain with high confidence ASD risk genes, and near FMRP gene targets are more likely to be in co-localized regions if they occur in ASD probands versus in their unaffected siblings. We also observed a similar enrichment for mutations near lincRNAs, previously shown to co-express with ASD risk genes. Additionally, we provide strong evidence that prioritized de novo mutations in autism probands point to a small set of well-known ASD genes, the disruption of which produces relevant mouse phenotypes such as abnormal social investigation and abnormal discrimination/associative learning, unlike the de novo mutations in unaffected siblings. The genome-wide co-localization results are available online.

Integrating Peak Colocalization and Motif Enrichment Analysis for the Discovery of Genome-Wide Regulatory Modules and Transcription Factor Recruitment Rules

Chromatin immunoprecipitation followed by next-generation sequencing (ChIP-Seq) has opened new avenues of research in the genome-wide characterization of regulatory DNA-protein interactions at the genetic and epigenetic level. As a consequence, it has become the de facto standard for studies on the regulation of transcription, and literally thousands of data sets for transcription factors and cofactors in different conditions and species are now available to the scientific community. However, while pipelines and best practices have been established for the analysis of a single experiment, there is still no consensus on the best way to perform an integrated analysis of multiple datasets in the same condition, in order to identify the most relevant and widespread regulatory modules composed by different transcription factors and cofactors. We present here a computational pipeline for this task, that integrates peak summit colocalization, a novel statistical framework for the evaluation of its significance, and motif enrichment analysis. We show examples of its application to ENCODE data, that led to the identification of relevant regulatory modules composed of different factors, as well as the organization on DNA of the binding motifs responsible for their recruitment.

epiCOLOC: Integrating Large-Scale and Context-Dependent Epigenomics Features for Comprehensive Colocalization Analysis

High-throughput genome-wide epigenomic assays, such as ChIP-seq, DNase-seq and ATAC-seq, have profiled a huge number of functional elements across numerous human tissues/cell types, which provide an unprecedented opportunity to interpret human genome and disease in context-dependent manner. Colocalization analysis determines whether genomic features are functionally related to a given search and will facilitate identifying the underlying biological functions characterizing intricate relationships with queries for genomic regions. Existing colocalization methods leveraged diverse assumptions and background models to assess the significance of enrichment, however, they only provided limited and predefined sets of epigenomic features. Here, we comprehensively collected and integrated over 44,385 bulk or single-cell epigenomic assays across 53 human tissues/cell types, such as transcription factor binding, histone modification, open chromatin and transcriptional event. By classifying these profiles into hierarchy of tissue/cell type, we developed a web portal, epiCOLOC (http://mulinlab.org/epicoloc or http://mulinlab.tmu.edu.cn/epicoloc), for users to perform context-dependent colocalization analysis in a convenient way.

A transcriptome-wide Mendelian randomization study to uncover tissue-dependent regulatory mechanisms across the human phenome

Developing insight into tissue-specific transcriptional mechanisms can help improve our understanding of how genetic variants exert their effects on complex traits and disease. In this study, we apply the principles of Mendelian randomization to systematically evaluate transcriptome-wide associations between gene expression (across 48 different tissue types) and 395 complex traits. Our findings indicate that variants which influence gene expression levels in multiple tissues are more likely to influence multiple complex traits. Moreover, detailed investigations of our results highlight tissue-specific associations, drug validation opportunities, insight into the likely causal pathways for trait-associated variants and also implicate putative associations at loci yet to be implicated in disease susceptibility. Similar evaluations can be conducted at http://mrcieu.mrsoftware.org/Tissue_MR_atlas/.

The Functional 3D Organization of Unicellular Genomes

Genome conformation capture techniques permit a systematic investigation into the functional spatial organization of genomes, including functional aspects like assessing the co-localization of sets of genomic elements. For example, the co-localization of genes targeted by a transcription factor (TF) within a transcription factory. We quantify spatial co-localization using a rigorous statistical model that measures the enrichment of a subset of elements in neighbourhoods inferred from Hi-C data. We also control for co-localization that can be attributed to genomic order. We systematically apply our open-sourced framework, spatial-mHG, to search for spatial co-localization phenomena in multiple unicellular Hi-C datasets with corresponding genomic annotations. Our biological findings shed new light on the functional spatial organization of genomes, including: In C. crescentus, DNA replication genes reside in two genomic clusters that are spatially co-localized. Furthermore, these clusters contain similar gene copies and lay in genomic vicinity to the ori and ter sequences. In S. cerevisae, Ty5 retrotransposon family element spatially co-localize at a spatially adjacent subset of telomeres. In N. crassa, both Proteasome lid subcomplex genes and protein refolding genes jointly spatially co-localize at a shared location. An implementation of our algorithms is available online.