eQTL Catalogue 2023: New datasets, X chromosome QTLs, and improved detection and visualisation of transcript-level QTLs

Variant-specific priors clarify colocalisation analysis

Linking GWAS variants to their causal gene and context remains an ongoing challenge. A widely used method for performing this analysis is the coloc package for statistical colocalisation analysis, which can be used to link GWAS and eQTL associations. Currently, coloc assumes that all variants in a region are equally likely to be causal, despite the success of fine-mapping methods that use additional information to adjust their prior probabilities. In this paper we propose and implement an approach for specifying variant-specific prior probabilities in the coloc method. We describe and compare six source of information for specifying prior probabilities: non-coding constraint, enhancer-gene link scores, the output of the PolyFun method and three estimates of eQTL-TSS distance densities. Using simulations and analysis of ground-truth pQTL-eQTL colocalisations we show that variant-specific priors, particularly the eQTL-TSS distance density priors, can improve colocalisation performance. Furthermore, across GWAS-eQTL colocalisations variant-specific priors changed colocalisation significance in up to 14.1% of colocalisations, at some loci revealing the likely causal gene.

Building a human genetic data lake to scale up insights for drug discovery

Genome-wide association studies (GWAS) have identified numerous disease-associated variants, yet efficient storage and analysis of genetic data remain a challenge. Here, we propose a scalable genetic data lake (GDL) integrating GWAS, molecular quantitative trait loci (mQTL), and epigenetic data within a big data infrastructure to enable rapid analysis. This framework allows large-scale computations, prioritizing 54 586 gene-trait associations, including 34 779 found exclusively in consortium data sets. By leveraging public, consortium, and private data, this approach enhances target discovery and indication selection, accelerating drug development.

Genome-wide association study reveals a novel tuberculosis susceptibility locus in multiple East Asian and European populations

Background

Tuberculosis (TB) continues to be a leading cause of morbidity and mortality worldwide. Past genome-wide association studies (GWAS) have explored TB susceptibility across various ethnic groups, yet a significant portion of TB heritability remains unexplained.

Methods

We conducted GWAS in the Singapore Chinese and Vietnamese, followed by a comprehensive meta-analysis incorporating 4 independent East Asian datasets, resulting in a total of 11,841 cases and 197,373 population controls.

Findings

We identified a novel susceptibility locus for pulmonary TB (PTB) at 22q12.2 in East Asians [rs6006426, OR (95%Cl) =1.097(1.066, 1.130), P meta =3.31×10 -10 ]. The association was further validated in Europeans [OR (95%Cl) =1.101(1.002, 1.211), P =0.046] and was strengthened in the combined meta-anlaysis including 12,736 PTB cases and 673,864 controls [OR (95%Cl) =1.098(1.068, 1.129), P meta =4.33×10 -11 ]. rs6006426 affected SF3A1 expression in various immune cells ( P from 0.003 to 6.17×10 -18 ) and OSM expression in monocytes post lipopolysaccharide stimulation ( P =5.57×10 -4 ). CRISPR-Cas9 edited zebrafish embryos with osm depletion resulted in decreased burden of Mycobacterium marinum ( M.marinum ) in infected embryos ( P =0.047).

Interpretation

Our findings offer novel insights into the genetic factors underlying TB and reveals new avenues for understanding its etiology.

Genome-wide association study for circulating metabolic traits in 619,372 individuals

Interpreting genetic associations with complex traits can be greatly improved by detailed understanding of the molecular consequences of these variants. However, although genome-wide association studies (GWAS) for common complex diseases routinely profile 1M+ individuals, studies of molecular phenotypes have lagged behind. We performed a GWAS meta-analysis for 249 circulating metabolic traits in the Estonian Biobank and the UK Biobank in up to 619,372 individuals, identifying 88,604 significant locus-metabolite associations and 8,774 independent lead variants, including 987 lead variants with a minor allele frequency less than 1%. We demonstrate how common and low-frequency associations converge on shared genes and pathways, bridging the gap between rare-variant burden testing and common-variant GWAS. We used Mendelian randomisation (MR) to explore putative causal links between metabolic traits, coronary artery disease and type 2 diabetes (T2D). Surprisingly, up to 85% of the tested metabolite-disease pairs had statistically significant genome-wide MR estimates, likely reflecting complex indirect effects driven by horisontal pleiotropy. To avoid these pleiotropic effects, we used cis-MR to test the phenotypic impact of inhibiting specific drug targets. We found that although plasma levels of branched-chain amino acids (BCAAs) have been associated with T2D in both observational and genome-wide MR studies, inhibiting the BCAA catabolism pathway to lower BCAA levels is unlikely to reduce T2D risk. Our publicly available results provide a valuable novel resource for GWAS interpretation and drug target prioritisation.

Transcriptome-wide analyses delineate the genetic architecture of expression variation in atopic dermatitis

Genome-wide association studies (GWASs) for atopic dermatitis (AD) have uncovered 81 risk loci in European participants; however, translating these findings into functional and therapeutic insights remains challenging. We conducted a transcriptome-wide association study (TWAS) in AD leveraging cis-eQTL data from sun exposed (n = 517), non-sun exposed skin (n = 602) and whole blood (n = 670) tissues and the latest GWAS of AD in Europeans (n = 864982). We implemented the OTTERS pipeline that combines polygenic risk score (PRS) techniques accommodating diverse assumptions in the architecture of gene regulation. We also used differential expression meta-analysis and co-expression networks (n = 186) to characterize the transcriptomic landscape of AD. We identified 176 gene-tissue associations covering 126 unique genes (53 previously unreported). Most TWAS risk genes were identified by adaptive PRS frameworks, with non-significant differences compared with clumping and thresholding approaches. TWAS risk genes were enriched in allergic reactions (e.g., AQP7, AFF4), skin barrier integrity (e.g., ACER3), and inflammatory pathways (e.g., TAPBPL). By integrating co-expression networks of lesional AD skin, we identified 16 hub genes previously identified as TWAS risk genes (six previously unreported) that orchestrate inflammatory responses (e.g., HSPA4) and keratinization (e.g., LCE3E, LCE3D), serving as potential drug targets through drug-gene interactions. Consistent associations between all analyses were reported for FOSL1 and RORC. Collectively, our findings provide additional risk genes for AD with potential implications in therapeutic approaches.

Long-read RNA sequencing atlas of human microglia isoforms elucidates disease-associated genetic regulation of splicing

Microglia, the innate immune cells of the central nervous system, have been genetically implicated in multiple neurodegenerative diseases. Mapping the genetics of gene expression in human microglia has identified several loci associated with disease-associated genetic variants in microglia-specific regulatory elements. However, identifying genetic effects on splicing is challenging because of the use of short sequencing reads. Here, we present the isoform-centric microglia genomic atlas (isoMiGA), which leverages long-read RNA sequencing to identify 35,879 novel microglia isoforms. We show that these isoforms are involved in stimulation response and brain region specificity. We then quantified the expression of both known and novel isoforms in a multi-ancestry meta-analysis of 555 human microglia short-read RNA sequencing samples from 391 donors, and found associations with genetic risk loci in Alzheimer's and Parkinson's disease. We nominate several loci that may act through complex changes in isoform and splice-site usage.

Comparative Genomics and Epigenomics of Transcriptional Regulation

Transcriptional regulation in response to diverse physiological cues involves complicated biological processes. Recent initiatives that leverage whole genome sequencing and annotation of regulatory elements significantly contribute to our understanding of transcriptional gene regulation. Advances in the data sets available for comparative genomics and epigenomics can identify evolutionarily constrained regulatory variants and shed light on noncoding elements that influence transcription in different tissues and developmental stages across species. Most epigenomic data, however, are generated from healthy subjects at specific developmental stages. To bridge the genotype-phenotype gap, future research should focus on generating multidimensional epigenomic data under diverse physiological conditions. Farm animal species offer advantages in terms of feasibility, cost, and experimental design for such integrative analyses in comparison to humans. Deep learning modeling and cutting-edge technologies in sequencing and functional screening and validation also provide great promise for better understanding transcriptional regulation in this dynamic field.

ChromBPNet: bias factorized, base-resolution deep learning models of chromatin accessibility reveal cis-regulatory sequence syntax, transcription factor footprints and regulatory variants

Despite extensive mapping of cis-regulatory elements (cREs) across cellular contexts with chromatin accessibility assays, the sequence syntax and genetic variants that regulate transcription factor (TF) binding and chromatin accessibility at context-specific cREs remain elusive. We introduce ChromBPNet, a deep learning DNA sequence model of base-resolution accessibility profiles that detects, learns and deconvolves assay-specific enzyme biases from regulatory sequence determinants of accessibility, enabling robust discovery of compact TF motif lexicons, cooperative motif syntax and precision footprints across assays and sequencing depths. Extensive benchmarks show that ChromBPNet, despite its lightweight design, is competitive with much larger contemporary models at predicting variant effects on chromatin accessibility, pioneer TF binding and reporter activity across assays, cell contexts and ancestry, while providing interpretation of disrupted regulatory syntax. ChromBPNet also helps prioritize and interpret regulatory variants that influence complex traits and rare diseases, thereby providing a powerful lens to decode regulatory DNA and genetic variation.

A bioinformatics toolbox to prioritize causal genetic variants in candidate regions

This review addresses the significant challenge of identifying causal genetic variants within quantitative trait loci (QTLs) for complex traits and diseases. Despite progress in detecting the ever-larger number of such loci, establishing causality remains daunting. We advocate for integrating bioinformatics and multiomics analyses to streamline the prioritization of candidate genes' variants. Our case study on the Pla2g4e gene, identified previously as a positional candidate obesity gene through genetic mapping and expression studies, demonstrates how applying multiomic data filtered through regulatory elements containing SNPs can refine the search for causative variants. This approach can yield results that guide more efficient experimental strategies, accelerating genetic research toward functional validation and therapeutic development.

Multimodal analysis of RNA sequencing data powers discovery of complex trait genetics

RNA sequencing has the potential to reveal many modalities of transcriptional regulation, such as various splicing phenotypes, but studies on gene regulation are often limited to gene expression due to the complexity of extracting and analyzing multiple RNA phenotypes. Here, we present Pantry, a framework to efficiently generate diverse RNA phenotypes from RNA sequencing data and perform downstream integrative analyses with genetic data. Pantry generates phenotypes from six modalities of transcriptional regulation (gene expression, isoform ratios, splice junction usage, alternative TSS/polyA usage, and RNA stability) and integrates them with genetic data via QTL mapping, TWAS, and colocalization testing. We apply Pantry to Geuvadis and GTEx data, finding that 4768 of the genes with no identified eQTL in Geuvadis have QTL in at least one other transcriptional modality, resulting in a 66% increase in genes over eQTL mapping. We further found that the QTL exhibit modality-specific functional properties that are further reinforced by joint analysis of different RNA modalities. We also show that generalizing TWAS to multiple RNA modalities approximately doubles the discovery of unique gene-trait associations, and enhances identification of regulatory mechanisms underlying GWAS signal in 42% of previously associated gene-trait pairs.

A brief guide to analyzing expression quantitative trait loci

Molecular quantitative trait locus (molQTL) mapping has emerged as an important approach for elucidating the functional consequences of genetic variants and unraveling the causal mechanisms underlying diseases or complex traits. However, the variety of analysis tools and sophisticated methodologies available for molQTL studies can be overwhelming for researchers with limited computational expertise. Here, we provide a brief guideline with a curated list of methods and software tools for analyzing expression quantitative trait loci, the most widely studied type of molQTL.

Extensive co-regulation of neighboring genes complicates the use of eQTLs in target gene prioritization

Identifying causal genes underlying genome-wide association studies (GWASs) is a fundamental problem in human genetics. Although colocalization with gene expression quantitative trait loci (eQTLs) is often used to prioritize GWAS target genes, systematic benchmarking has been limited due to unavailability of large ground truth datasets. Here, we re-analyzed plasma protein QTL data from 3,301 individuals of the INTERVAL cohort together with 131 eQTL Catalog datasets. Focusing on variants located within or close to the affected protein identified 793 proteins with at least one cis-pQTL where we could assume that the most likely causal gene was the gene coding for the protein. We then benchmarked the ability of cis-eQTLs to recover these causal genes by comparing three Bayesian colocalization methods (coloc.susie, coloc.abf, and CLPP) and five Mendelian randomization (MR) approaches (three varieties of inverse-variance weighted MR, MR-RAPS, and MRLocus). We found that assigning fine-mapped pQTLs to their closest protein coding genes outperformed all colocalization methods regarding both precision (71.9%) and recall (76.9%). Furthermore, the colocalization method with the highest recall (coloc.susie - 46.3%) also had the lowest precision (45.1%). Combining evidence from multiple conditionally distinct colocalizing QTLs with MR increased precision to 81%, but this was accompanied by a large reduction in recall to 7.1%. Furthermore, the choice of the MR method greatly affected performance, with the standard inverse-variance-weighted MR often producing many false positives. Our results highlight that linking GWAS variants to target genes remains challenging with eQTL evidence alone, and prioritizing novel targets requires triangulation of evidence from multiple sources.

A compendium of genetic variations associated with promoter usage across 49 human tissues

Promoters play a crucial role in regulating gene transcription. However, our understanding of how genetic variants influence alternative promoter selection is still incomplete. In this study, we implement a framework to identify genetic variants that affect the relative usage of alternative promoters, known as promoter usage quantitative trait loci (puQTLs). By constructing an atlas of human puQTLs across 49 different tissues from 838 individuals, we have identified approximately 76,856 independent loci associated with promoter usage, encompassing 602,009 genetic variants. Our study demonstrates that puQTLs represent a distinct type of molecular quantitative trait loci, effectively uncovering regulatory targets and patterns. Furthermore, puQTLs are regulating in a tissue-specific manner and are enriched with binding sites of epigenetic marks and transcription factors, especially those involved in chromatin architecture formation. Notably, we have also found that puQTLs colocalize with complex traits or diseases and contribute to their heritability. Collectively, our findings underscore the significant role of puQTLs in elucidating the molecular mechanisms underlying tissue development and complex diseases.

Unlocking the genetic influence on milk variation and its potential implication for infant health

Human milk has long been recognized for its critical role in infant and maternal health. In this issue of Cell Genomics, Johnson et al.1 apply a human genetics and genomics approach to shed light on the complex relationship between maternal genetics, milk variation, and the infant gut microbiome.

Long-read proteogenomics to connect disease-associated sQTLs to the protein isoform effectors of disease

A major fraction of loci identified by genome-wide association studies (GWASs) mediate alternative splicing, but mechanistic interpretation is hindered by the technical limitations of short-read RNA sequencing (RNA-seq), which cannot directly link splicing events to full-length protein isoforms. Long-read RNA-seq represents a powerful tool to characterize transcript isoforms, and recently, infer protein isoform existence. Here, we present an approach that integrates information from GWASs, splicing quantitative trait loci (sQTLs), and PacBio long-read RNA-seq in a disease-relevant model to infer the effects of sQTLs on the ultimate protein isoform products they encode. We demonstrate the utility of our approach using bone mineral density (BMD) GWAS data. We identified 1,863 sQTLs from the Genotype-Tissue Expression (GTEx) project in 732 protein-coding genes that colocalized with BMD associations (H4PP ≥ 0.75). We generated PacBio Iso-Seq data (N = ∼22 million full-length reads) on human osteoblasts, identifying 68,326 protein-coding isoforms, of which 17,375 (25%) were unannotated. By casting the sQTLs onto protein isoforms, we connected 809 sQTLs to 2,029 protein isoforms from 441 genes expressed in osteoblasts. Overall, we found that 74 sQTLs influenced isoforms likely impacted by nonsense-mediated decay and 190 that potentially resulted in the expression of unannotated protein isoforms. Finally, we functionally validated colocalizing sQTLs in TPM2, in which siRNA-mediated knockdown in osteoblasts showed two TPM2 isoforms with opposing effects on mineralization but exhibited no effect upon knockdown of the entire gene. Our approach should be to generalize across diverse clinical traits and to provide insights into protein isoform activities modulated by GWAS loci.

Metadata-guided feature disentanglement for functional genomics

With the development of high-throughput technologies, genomics datasets rapidly grow in size, including functional genomics data. This has allowed the training of large Deep Learning (DL) models to predict epigenetic readouts, such as protein binding or histone modifications, from genome sequences. However, large dataset sizes come at a price of data consistency, often aggregating results from a large number of studies, conducted under varying experimental conditions. While data from large-scale consortia are useful as they allow studying the effects of different biological conditions, they can also contain unwanted biases from confounding experimental factors. Here, we introduce Metadata-guided Feature Disentanglement (MFD)-an approach that allows disentangling biologically relevant features from potential technical biases. MFD incorporates target metadata into model training, by conditioning weights of the model output layer on different experimental factors. It then separates the factors into disjoint groups and enforces independence of the corresponding feature subspaces with an adversarially learned penalty. We show that the metadata-driven disentanglement approach allows for better model introspection, by connecting latent features to experimental factors, without compromising, or even improving performance in downstream tasks, such as enhancer prediction, or genetic variant discovery. The code will be made available at https://github.com/HealthML/MFD.

Genetically proxied HTRA1 protease activity and circulating levels independently predict risk of ischemic stroke and coronary artery disease

Genetic variants in HTRA1 are associated with stroke risk. However, the mechanisms mediating this remain largely unknown, as does the full spectrum of phenotypes associated with genetic variation in HTRA1. Here we show that rare HTRA1 variants are linked to ischemic stroke in the UK Biobank and BioBank Japan. Integrating data from biochemical experiments, we next show that variants causing loss of protease function associated with ischemic stroke, coronary artery disease and skeletal traits in the UK Biobank and MyCode cohorts. Moreover, a common variant modulating circulating HTRA1 mRNA and protein levels enhances the risk of ischemic stroke and coronary artery disease while lowering the risk of migraine and macular dystrophy in genome-wide association study, UK Biobank, MyCode and BioBank Japan data. We found no interaction between proxied HTRA1 activity and levels. Our findings demonstrate the role of HTRA1 for cardiovascular diseases and identify two mechanisms as potential targets for therapeutic interventions.

Multimodal analysis of RNA sequencing data powers discovery of complex trait genetics

Transcriptome data is commonly used to understand genome function via quantitative trait loci (QTL) mapping and to identify the molecular mechanisms driving genome wide association study (GWAS) signals through colocalization analysis and transcriptome-wide association studies (TWAS). While RNA sequencing (RNA-seq) has the potential to reveal many modalities of transcriptional regulation, such as various splicing phenotypes, such studies are often limited to gene expression due to the complexity of extracting and analyzing multiple RNA phenotypes. Here, we present Pantry (Pan-transcriptomic phenotyping), a framework to efficiently generate diverse RNA phenotypes from RNA-seq data and perform downstream integrative analyses with genetic data. Pantry currently generates phenotypes from six modalities of transcriptional regulation (gene expression, isoform ratios, splice junction usage, alternative TSS/polyA usage, and RNA stability) and integrates them with genetic data via QTL mapping, TWAS, and colocalization testing. We applied Pantry to Geuvadis and GTEx data, and found that 4,768 of the genes with no identified expression QTL in Geuvadis had QTLs in at least one other transcriptional modality, resulting in a 66% increase in genes over expression QTL mapping. We further found that QTLs exhibit modality-specific functional properties that are further reinforced by joint analysis of different RNA modalities. We also show that generalizing TWAS to multiple RNA modalities (xTWAS) approximately doubles the discovery of unique gene-trait associations, and enhances identification of regulatory mechanisms underlying GWAS signal in 42% of previously associated gene-trait pairs. We provide the Pantry code, RNA phenotypes from all Geuvadis and GTEx samples, and xQTL and xTWAS results on the web.