The EN-TEx resource of multi-tissue personal epigenomes & variant-impact models

Addressing Technical Pitfalls in Pursuit of Molecular Factors That Mediate Immunoglobulin Gene Regulation

The expressed Ab repertoire is a critical determinant of immune-related phenotypes. Ab-encoding transcripts are distinct from other expressed genes because they are transcribed from somatically rearranged gene segments. Human Abs are composed of two identical H and L chain polypeptides derived from genes in IGH locus and one of two L chain loci. The combinatorial diversity that results from Ab gene rearrangement and the pairing of different H and L chains contributes to the immense diversity of the baseline Ab repertoire. During rearrangement, Ab gene selection is mediated by factors that influence chromatin architecture, promoter/enhancer activity, and V(D)J recombination. Interindividual variation in the composition of the Ab repertoire associates with germline variation in IGH, implicating polymorphism in Ab gene regulation. Determining how IGH variants directly mediate gene regulation will require integration of these variants with other functional genomic datasets. In this study, we argue that standard approaches using short reads have limited utility for characterizing regulatory regions in IGH at haplotype resolution. Using simulated and chromatin immunoprecipitation sequencing reads, we define features of IGH that limit use of short reads and a single reference genome, namely 1) the highly duplicated nature of the DNA sequence in IGH and 2) structural polymorphisms that are frequent in the population. We demonstrate that personalized diploid references enhance performance of short-read data for characterizing mappable portions of the locus, while also showing that long-read profiling tools will ultimately be needed to fully resolve functional impacts of IGH germline variation on expressed Ab repertoires.

Human genetic variation determines 24-hour rhythmic gene expression and disease risk

24-hour biological rhythms are essential to maintain physiological homeostasis. Disruption of these rhythms increases the risks of multiple diseases. The biological rhythms are known to have a genetic basis formed by core clock genes, but how individual genetic variation shapes the oscillating transcriptome and contributes to human chronophysiology and disease risk is largely unknown. Here, we mapped interactions between temporal gene expression and genotype to identify quantitative trait loci (QTLs) contributing to rhythmic gene expression. These newly identified QTLs were termed as rhythmic QTLs (rhyQTLs), which determine previously unappreciated rhythmic genes in human subpopulations with specific genotypes. Functionally, rhyQTLs and their associated rhythmic genes contribute extensively to essential chronophysiological processes, including bile acid and lipid metabolism. The identification of rhyQTLs sheds light on the genetic mechanisms of gene rhythmicity, offers mechanistic insights into variations in human disease risk, and enables precision chronotherapeutic approaches for patients.

Big data and deep learning for RNA biology

The exponential growth of big data in RNA biology (RB) has led to the development of deep learning (DL) models that have driven crucial discoveries. As constantly evidenced by DL studies in other fields, the successful implementation of DL in RB depends heavily on the effective utilization of large-scale datasets from public databases. In achieving this goal, data encoding methods, learning algorithms, and techniques that align well with biological domain knowledge have played pivotal roles. In this review, we provide guiding principles for applying these DL concepts to various problems in RB by demonstrating successful examples and associated methodologies. We also discuss the remaining challenges in developing DL models for RB and suggest strategies to overcome these challenges. Overall, this review aims to illuminate the compelling potential of DL for RB and ways to apply this powerful technology to investigate the intriguing biology of RNA more effectively.

Single-cell genomics and regulatory networks for 388 human brains

Single-cell genomics is a powerful tool for studying heterogeneous tissues such as the brain. Yet little is understood about how genetic variants influence cell-level gene expression. Addressing this, we uniformly processed single-nuclei, multiomics datasets into a resource comprising >2.8 million nuclei from the prefrontal cortex across 388 individuals. For 28 cell types, we assessed population-level variation in expression and chromatin across gene families and drug targets. We identified >550,000 cell type-specific regulatory elements and >1.4 million single-cell expression quantitative trait loci, which we used to build cell-type regulatory and cell-to-cell communication networks. These networks manifest cellular changes in aging and neuropsychiatric disorders. We further constructed an integrative model accurately imputing single-cell expression and simulating perturbations; the model prioritized ~250 disease-risk genes and drug targets with associated cell types.

Haplotype function score improves biological interpretation and cross-ancestry polygenic prediction of human complex traits

We propose a new framework for human genetic association studies: at each locus, a deep learning model (in this study, Sei) is used to calculate the functional genomic activity score for two haplotypes per individual. This score, defined as the Haplotype Function Score (HFS), replaces the original genotype in association studies. Applying the HFS framework to 14 complex traits in the UK Biobank, we identified 3619 independent HFS-trait associations with a significance of p < 5 × 10-8. Fine-mapping revealed 2699 causal associations, corresponding to a median increase of 63 causal findings per trait compared with single-nucleotide polymorphism (SNP)-based analysis. HFS-based enrichment analysis uncovered 727 pathway-trait associations and 153 tissue-trait associations with strong biological interpretability, including 'circadian pathway-chronotype' and 'arachidonic acid-intelligence'. Lastly, we applied least absolute shrinkage and selection operator (LASSO) regression to integrate HFS prediction score with SNP-based polygenic risk scores, which showed an improvement of 16.1-39.8% in cross-ancestry polygenic prediction. We concluded that HFS is a promising strategy for understanding the genetic basis of human complex traits.

Multi-omic QTL mapping in early developmental tissues reveals phenotypic and temporal complexity of regulatory variants underlying GWAS loci

Most GWAS loci are presumed to affect gene regulation, however, only ∼43% colocalize with expression quantitative trait loci (eQTLs). To address this colocalization gap, we identify eQTLs, chromatin accessibility QTLs (caQTLs), and histone acetylation QTLs (haQTLs) using molecular samples from three early developmental (EDev) tissues. Through colocalization, we annotate 586 GWAS loci for 17 traits by QTL complexity, QTL phenotype, and QTL temporal specificity. We show that GWAS loci are highly enriched for colocalization with complex QTL modules that affect multiple elements (genes and/or peaks). We also demonstrate that caQTLs and haQTLs capture regulatory variations not associated with eQTLs and explain ∼49% of the functionally annotated GWAS loci. Additionally, we show that EDev-unique QTLs are strongly depleted for colocalizing with GWAS loci. By conducting one of the largest multi-omic QTL studies to date, we demonstrate that many GWAS loci exhibit phenotypic complexity and therefore, are missed by traditional eQTL analyses.

Epigenomic insights into common human disease pathology

The epigenome-the chemical modifications and chromatin-related packaging of the genome-enables the same genetic template to be activated or repressed in different cellular settings. This multi-layered mechanism facilitates cell-type specific function by setting the local sequence and 3D interactive activity level. Gene transcription is further modulated through the interplay with transcription factors and co-regulators. The human body requires this epigenomic apparatus to be precisely installed throughout development and then adequately maintained during the lifespan. The causal role of the epigenome in human pathology, beyond imprinting disorders and specific tumour suppressor genes, was further brought into the spotlight by large-scale sequencing projects identifying that mutations in epigenomic machinery genes could be critical drivers in both cancer and developmental disorders. Abrogation of this cellular mechanism is providing new molecular insights into pathogenesis. However, deciphering the full breadth and implications of these epigenomic changes remains challenging. Knowledge is accruing regarding disease mechanisms and clinical biomarkers, through pathogenically relevant and surrogate tissue analyses, respectively. Advances include consortia generated cell-type specific reference epigenomes, high-throughput DNA methylome association studies, as well as insights into ageing-related diseases from biological 'clocks' constructed by machine learning algorithms. Also, 3rd-generation sequencing is beginning to disentangle the complexity of genetic and DNA modification haplotypes. Cell-free DNA methylation as a cancer biomarker has clear clinical utility and further potential to assess organ damage across many disorders. Finally, molecular understanding of disease aetiology brings with it the opportunity for exact therapeutic alteration of the epigenome through CRISPR-activation or inhibition.

Dissection of a CTCF topological boundary uncovers principles of enhancer-oncogene regulation

Enhancer-gene communication is dependent on topologically associating domains (TADs) and boundaries enforced by the CCCTC-binding factor (CTCF) insulator, but the underlying structures and mechanisms remain controversial. Here, we investigate a boundary that typically insulates fibroblast growth factor (FGF) oncogenes but is disrupted by DNA hypermethylation in gastrointestinal stromal tumors (GISTs). The boundary contains an array of CTCF sites that enforce adjacent TADs, one containing FGF genes and the other containing ANO1 and its putative enhancers, which are specifically active in GIST and its likely cell of origin. We show that coordinate disruption of four CTCF motifs in the boundary fuses the adjacent TADs, allows the ANO1 enhancer to contact FGF3, and causes its robust induction. High-resolution micro-C maps reveal specific contact between transcription initiation sites in the ANO1 enhancer and FGF3 promoter that quantitatively scales with FGF3 induction such that modest changes in contact frequency result in strong changes in expression, consistent with a causal relationship.

Single-cell genomics and regulatory networks for 388 human brains

Single-cell genomics is a powerful tool for studying heterogeneous tissues such as the brain. Yet, little is understood about how genetic variants influence cell-level gene expression. Addressing this, we uniformly processed single-nuclei, multi-omics datasets into a resource comprising >2.8M nuclei from the prefrontal cortex across 388 individuals. For 28 cell types, we assessed population-level variation in expression and chromatin across gene families and drug targets. We identified >550K cell-type-specific regulatory elements and >1.4M single-cell expression-quantitative-trait loci, which we used to build cell-type regulatory and cell-to-cell communication networks. These networks manifest cellular changes in aging and neuropsychiatric disorders. We further constructed an integrative model accurately imputing single-cell expression and simulating perturbations; the model prioritized ~250 disease-risk genes and drug targets with associated cell types.

Small polymorphisms are a source of ancestral bias in structural variant breakpoint placement

High-quality genome assemblies and sophisticated algorithms have increased sensitivity for a wide range of variant types, and breakpoint accuracy for structural variants (SVs, ≥50 bp) has improved to near base pair precision. Despite these advances, many SV breakpoint locations are subject to systematic bias affecting variant representation. To understand why SV breakpoints are inconsistent across samples, we reanalyzed 64 phased haplotypes constructed from long-read assemblies released by the Human Genome Structural Variation Consortium (HGSVC). We identify 882 SV insertions and 180 SV deletions with variable breakpoints not anchored in tandem repeats (TRs) or segmental duplications (SDs). SVs called from aligned sequencing reads increase breakpoint disagreements by 2×-16×. Sequence accuracy had a minimal impact on breakpoints, but we observe a strong effect of ancestry. We confirm that SNP and indel polymorphisms are enriched at shifted breakpoints and are also absent from variant callsets. Breakpoint homology increases the likelihood of imprecise SV calls and the distance they are shifted, and tandem duplications are the most heavily affected SVs. Because graph genome methods normalize SV calls across samples, we investigated graphs generated by two different methods and find the resulting breakpoints are subject to other technical biases affecting breakpoint accuracy. The breakpoint inconsistencies we characterize affect ∼5% of the SVs called in a human genome and can impact variant interpretation and annotation. These limitations underscore a need for algorithm development to improve SV databases, mitigate the impact of ancestry on breakpoints, and increase the value of callsets for investigating breakpoint features.

Comparative analysis of models in predicting the effects of SNPs on TF-DNA binding using large-scale in vitro and in vivo data

Non-coding variants associated with complex traits can alter the motifs of transcription factor (TF)-deoxyribonucleic acid binding. Although many computational models have been developed to predict the effects of non-coding variants on TF binding, their predictive power lacks systematic evaluation. Here we have evaluated 14 different models built on position weight matrices (PWMs), support vector machines, ordinary least squares and deep neural networks (DNNs), using large-scale in vitro (i.e. SNP-SELEX) and in vivo (i.e. allele-specific binding, ASB) TF binding data. Our results show that the accuracy of each model in predicting SNP effects in vitro significantly exceeds that achieved in vivo. For in vitro variant impact prediction, kmer/gkm-based machine learning methods (deltaSVM_HT-SELEX, QBiC-Pred) trained on in vitro datasets exhibit the best performance. For in vivo ASB variant prediction, DNN-based multitask models (DeepSEA, Sei, Enformer) trained on the ChIP-seq dataset exhibit relatively superior performance. Among the PWM-based methods, tRap demonstrates better performance in both in vitro and in vivo evaluations. In addition, we find that TF classes such as basic leucine zipper factors could be predicted more accurately, whereas those such as C2H2 zinc finger factors are predicted less accurately, aligning with the evolutionary conservation of these TF classes. We also underscore the significance of non-sequence factors such as cis-regulatory element type, TF expression, interactions and post-translational modifications in influencing the in vivo predictive performance of TFs. Our research provides valuable insights into selecting prioritization methods for non-coding variants and further optimizing such models.

Comprehensive and accurate genome analysis at scale using DRAGEN accelerated algorithms

Research and medical genomics require comprehensive and scalable solutions to drive the discovery of novel disease targets, evolutionary drivers, and genetic markers with clinical significance. This necessitates a framework to identify all types of variants independent of their size (e.g., SNV/SV) or location (e.g., repeats). Here we present DRAGEN that utilizes novel methods based on multigenomes, hardware acceleration, and machine learning based variant detection to provide novel insights into individual genomes with ~30min computation time (from raw reads to variant detection). DRAGEN outperforms all other state-of-the-art methods in speed and accuracy across all variant types (SNV, indel, STR, SV, CNV) and further incorporates specialized methods to obtain key insights in medically relevant genes (e.g., HLA, SMN, GBA). We showcase DRAGEN across 3,202 genomes and demonstrate its scalability, accuracy, and innovations to further advance the integration of comprehensive genomics for research and medical applications.

Characterizing prostate cancer risk through multi-ancestry genome-wide discovery of 187 novel risk variants

The transferability and clinical value of genetic risk scores (GRSs) across populations remain limited due to an imbalance in genetic studies across ancestrally diverse populations. Here we conducted a multi-ancestry genome-wide association study of 156,319 prostate cancer cases and 788,443 controls of European, African, Asian and Hispanic men, reflecting a 57% increase in the number of non-European cases over previous prostate cancer genome-wide association studies. We identified 187 novel risk variants for prostate cancer, increasing the total number of risk variants to 451. An externally replicated multi-ancestry GRS was associated with risk that ranged from 1.8 (per standard deviation) in African ancestry men to 2.2 in European ancestry men. The GRS was associated with a greater risk of aggressive versus non-aggressive disease in men of African ancestry (P = 0.03). Our study presents novel prostate cancer susceptibility loci and a GRS with effective risk stratification across ancestry groups.

Population variability in X-chromosome inactivation across 9 mammalian species

One of the two X chromosomes in female mammals is epigenetically silenced in embryonic stem cells by X chromosome inactivation (XCI). This creates a mosaic of cells expressing either the maternal or the paternal X allele. The XCI ratio, the proportion of inactivated parental alleles, varies widely among individuals, representing the largest instance of epigenetic variability within mammalian populations. While various contributing factors to XCI variability are recognized, namely stochastic and/or genetic effects, their relative contributions are poorly understood. This is due in part to limited cross-species analysis, making it difficult to distinguish between generalizable or species-specific mechanisms for XCI ratio variability. To address this gap, we measured XCI ratios in nine mammalian species (9,143 individual samples), ranging from rodents to primates, and compared the strength of stochastic models or genetic factors for explaining XCI variability. Our results demonstrate the embryonic stochasticity of XCI is a general explanatory model for population XCI variability in mammals, while genetic factors play a minor role.

Allele biased transcription factor binding across human brain regions gives mechanistic insight into eQTLs

Transcription Factors (TFs) influence gene expression by facilitating or disrupting the formation of transcription initiation machinery at particular genomic loci. Because genomic localization of TFs is in part driven by TF recognition of DNA sequence, variation in TF binding sites can disrupt TF-DNA associations and affect gene regulation. To identify variants that impact TF binding in human brain tissues, we quantified allele bias for 93 TFs analyzed with ChIP-seq experiments of multiple structural brain regions from two donors. Using graph genomes constructed from phased genomic sequence data, we compared ChIP-seq signal between alleles at heterozygous variants within each tissue sample from each donor. Comparison of results from different brain regions within donors and the same regions between donors provided measures of allele bias reproducibility. We identified thousands of DNA variants that show reproducible bias in ChIP-seq for at least one TF. We found that alleles that are rarer in the general population were more likely than common alleles to exhibit large biases, and more frequently led to reduced TF binding. Combining ChIP-seq with RNA-seq, we identified TF-allele interaction biases with RNA bias in a phased allele linked to 6,709 eQTL variants identified in GTEx data, 3,309 of which were found in neural contexts. Our results provide insights into the effects of both common and rare variation on gene regulation in the brain. These findings can facilitate mechanistic understanding of cis-regulatory variation associated with biological traits, including disease.

Getting personal with epigenetics: towards individual-specific epigenomic imputation with machine learning

Epigenetic modifications are dynamic mechanisms involved in the regulation of gene expression. Unlike the DNA sequence, epigenetic patterns vary not only between individuals, but also between different cell types within an individual. Environmental factors, somatic mutations and ageing contribute to epigenetic changes that may constitute early hallmarks or causal factors of disease. Epigenetic modifications are reversible and thus promising therapeutic targets for precision medicine. However, mapping efforts to determine an individual's cell-type-specific epigenome are constrained by experimental costs and tissue accessibility. To address these challenges, we developed eDICE, an attention-based deep learning model that is trained to impute missing epigenomic tracks by conditioning on observed tracks. Using a recently published set of epigenomes from four individual donors, we show that transfer learning across individuals allows eDICE to successfully predict individual-specific epigenetic variation even in tissues that are unmapped in a given donor. These results highlight the potential of machine learning-based imputation methods to advance personalized epigenomics.

Small allelic variants are a source of ancestral bias in structural variant breakpoint placement

High-quality genome assemblies and sophisticated algorithms have increased sensitivity for a wide range of variant types, and breakpoint accuracy for structural variants (SVs, ≥ 50 bp) has improved to near basepair precision. Despite these advances, many SVs in unique regions of the genome are subject to systematic bias that affects breakpoint location. This ambiguity leads to less accurate variant comparisons across samples, and it obscures true breakpoint features needed for mechanistic inferences. To understand why SVs are not consistently placed, we reanalyzed 64 phased haplotypes constructed from long-read assemblies released by the Human Genome Structural Variation Consortium (HGSVC). We identified variable breakpoints for 882 SV insertions and 180 SV deletions not anchored in tandem repeats (TRs) or segmental duplications (SDs). While this is unexpectedly high for genome assemblies in unique loci, we find read-based callsets from the same sequencing data yielded 1,566 insertions and 986 deletions with inconsistent breakpoints also not anchored in TRs or SDs. When we investigated causes for breakpoint inaccuracy, we found sequence and assembly errors had minimal impact, but we observed a strong effect of ancestry. We confirmed that polymorphic mismatches and small indels are enriched at shifted breakpoints and that these polymorphisms are generally lost when breakpoints shift. Long tracts of homology, such as SVs mediated by transposable elements, increase the likelihood of imprecise SV calls and the distance they are shifted. Tandem Duplication (TD) breakpoints are the most heavily affected SV class with 14% of TDs placed at different locations across haplotypes. While graph genome methods normalize SV calls across many samples, the resulting breakpoints are sometimes incorrect, highlighting a need to tune graph methods for breakpoint accuracy. The breakpoint inconsistencies we characterize collectively affect ~5% of the SVs called in a human genome and underscore a need for algorithm development to improve SV databases, mitigate the impact of ancestry on breakpoint placement, and increase the value of callsets for investigating mutational processes.