Biology subjects

Delaneau, O.

Publications and source records attributed to Delaneau, O..

At least 19 recordsLinked to original sources

bioRxiv2025

Benchmarking imputation accuracy in the presence or absence of a reference panel

Whole genome sequencing (WGS) of a large number of samples is costly. Solutions to reduce this cost include targeting a proportion of the genome (e.g. SNP arrays) or lowering the sequencing depth (low-coverage WGS, lcWGS) but both solutions suffer from either genotype missingness or uncertainty. Genomic imputation addresses this problem by inferring missing or uncertain genotypes using a collection of high quality genomic data (a reference panel). However, certain methods can impute a lcWGS dataset without a reference panel. Because the investment into generating a reference panel can be prohibitively expensive in non-model species, a benchmarking of the accuracy of these alternative methods of imputation can help inform study design. Here, we imputed a dataset of 2800 lcWGS in the presence and absence of a reference panel of 502 samples. We used 32 individuals sequenced at both high and low coverage to estimate the accuracy of each method and explored the limitations of lcWGS sample size and reference panel size. Although the best results were achieved with a large reference panel, using only a lcWGS dataset showed accurate imputation, when over 500 samples were used and we account for missing data and low frequency alleles. In addition, imputing with or without a reference panel gave similar results in a GWAS for a polygenic trait but required some quality control in the identification of homozygous-by-descent segments. Thus, while using a reference panel remains the ideal approach, imputation in suitably large lcWGS datasets can provide sufficient accuracy given proper quality control.

genomics↗

bioRxiv2025

Variant Classification Using Proteomics-Informed Large Language Models Increases Power of Rare Variant Association Studies and Enhances Target Discovery

Rare variant association analysis, which assesses the aggregate effect of rare damaging variants within a gene, is a powerful strategy for advancing knowledge of human biology. Numerous models have been proposed to identify damaging coding variants, with the most recent ones employing deep learning and large language models (LLM) to predict the impact of changes in coding sequences. Here, we use newly available proteomics data on 2,898 proteins across 46, 665 individuals to evaluate and refine LLM predictors of damaging variants. Using one of these refined models, we evaluate association between rare damaging variants and human phenotypes at 241 positive control gene-trait pairs. Among these gene-trait pairs, our proteomics-guided model outperforms an ensemble of conventional approaches including PolyPhen2, Mutation Taster, SIFT, and LRT, as well as newer machine learning approaches for identifying damaging missense variants, such as CADD, ESM-1v, ESM-1b and AlphaMissense. When attempting to recover known associations by correctly separating damaging singleton missense variants from other singleton variants, our approach recapitulates 36.5% of gene-trait pairs with known associations, exceeding all the alternatives we considered. Furthermore, when we apply our model to 10 exemplary traits from the UK Biobank, we identify 177 gene-trait associations - again exceeding all other approaches. Our results demonstrate that summary statistics from large-scale human proteomics data enable evaluation and refinement of coding variant classification LLMs, improving discovery potential in human genetic studies.

bioinformatics↗

bioRxiv2025

Separating direct, indirect and parent-of-origin genetic effects in the human population

Here, we present a novel approach to estimate the degree to which the phenotypic effect of a DNA locus is attributable to four components: alleles in the child (direct genetic effects), alleles in the mother and the father (indirect genetic effects), or is dependent upon the parent from which it is inherited (parent-of-origin, PofO effects). Applying our model, JODIE, to 30,000 child-mother-father trios with phased DNA information from the Estonian Biobank (EstBB) and the Norwegian Mother, Father, Child Cohort (MoBa), we jointly estimate the phenotypic variance attributable to these four effects unbiased of assortative mating (AM) for height, body mass index (BMI) and childhood educational test score (EA). For all three traits, direct effects make the largest contribution to the genetic effect variance. But we find that parental indirect genetic effects make an equivalent combined contribution, and that there is a non-zero PofO effect variance for all traits. We calculate the heritability that would be obtained at the populationlevel in the absence of AM for common DNA loci, and show that the proportional contribution of direct effects to these heritability values can be calculated as 64.0% for EA in MoBa, 77.1% and 63.4% for height in MoBa and EstBB, and 81.2% and 88.0% for BMI in MoBa and EstBB. Additionally, using within-family genome-wide association testing, we identify 276 independently associated DNA regions that replicate across two additional biobanks, which all show a genotype-phenotype relationship that reflects an interplay of direct, indirect and PofO effects. Determining how direct, parental and PofO genetic effects combine across loci genome-wide to influence human phenotypic variation requires joint modeling of parental and child genotypes alongside the parental origin of loci and here, we make the first attempt to do this in the human population.

genetics↗

bioRxiv2023

Assessing the impact of post-mortem damage and contamination on imputation performance in ancient DNA

Low-coverage imputation is becoming ever more present in ancient DNA (aDNA) studies. Imputation pipelines commonly used for present-day genomes have been shown to yield accurate results when applied to ancient genomes. However, post-mortem damage (PMD), in the form of C-to-T substitutions at the reads termini, and contamination with DNA from closely related species can potentially affect imputation performance in aDNA. In this study, we evaluated imputation performance i) when using a genotype caller designed for aDNA, ATLAS, compared to bcftools, and ii) when contamination is present. We evaluated imputation performance with principal component analyses (PCA) and by calculating imputation error rates. With a particular focus on differently imputed sites, we found that using ATLAS prior to imputation substantially improved imputed genotypes for a very damaged ancient genome (42% PMD). For the remaining genomes, ATLAS brought limited gains. Finally, to examine the effect of contamination on imputation, we added various amounts of reads from two present-day genomes to a previously downsampled high-coverage ancient genome. We observed that imputation accuracy drastically decreased for contamination rates above 5%. In conclusion, we recommend i) accounting for PMD by using a genotype caller such as ATLAS before imputing highly damaged genomes and ii) only imputing genomes containing up to 5% of contamination.

genomics↗

bioRxiv2023

Improving population scale statistical phasing with whole-genome sequencing data

Haplotype estimation, or phasing, has gained significant traction in large-scale projects due to its valuable contributions to population genetics, variant analysis, and the creation of reference panels for imputation and phasing of new samples. To scale with the growing number of samples, haplotype estimation methods designed for population scale rely on highly optimized statistical models to phase genotype data, and usually ignore read-level information. Statistical methods excel in resolving common variants, however, they still struggle at rare variants due to the lack of statistical information. In this study we introduce SAPPHIRE, a new method that leverages whole-genome sequencing data to enhance the precision of haplotype calls produced by statistical phasing. SAPPHIRE achieves this by refining haplotype estimates through the realignment of sequencing reads, particularly targeting low-confidence phase calls. Our findings demonstrate that SAPPHIRE significantly enhances the accuracy of haplotypes obtained from state of the art methods and also provides the subset of phase calls that are validated by sequencing reads. Finally, we show that our method scales to large data sets by its successful application to the extensive 3.6 Petabytes of sequencing data of the last UK Biobank 200,031 sample release.

genomics↗

bioRxiv2023

Non-coding rare variant associations with blood traits on 166,740 UK Biobank genomes

Large biobanks with whole-genome sequencing now enable the association of non-coding rare variants with complex human traits. Given that >98% of the genome is available for exploration, the selection of non-coding variants remains a critical yet unresolved challenge in these analyses. Here, we leverage knowledge of blood gene regulation and deleteriousness scores to select non-coding variants pertinent for association with blood-related traits. We leverage whole genome sequencing and 59 blood cell count and biomarker measurements for 166 740 UK Biobank samples to perform variant collapsing tests. We identified hundreds of gene-trait associations involving non-coding variants across the 59 traits. However, we demonstrate that the majority of these non-coding rare variant associations (i) reproduce associations known from common variant studies and (ii) are driven by linkage disequilibrium between nearby common and rare variants. This study underscores the prevailing challenges in rare variant analysis and the need for caution when interpreting non-coding rare variant association results.

genetics↗

bioRxiv2023

Polygenic risk score prediction accuracy convergence

Polygenic risk scores (PRS) trained from genome-wide association study (GWAS) results are set to play a pivotal role in biomedical research addressing multifactorial human diseases. The prospect of using these risk scores in clinical care and public health is generating both enthusiasm and controversy, with varying opinions about strengths and limitations across experts1. The performances of existing polygenic scores are still limited, and although it is expected to improve with increasing sample size of GWAS and the development of new powerful methods, it remains unclear how much prediction can be ultimately achieved. Here, we conducted a retrospective analysis to assess the progress in PRS prediction accuracy since the publication of the first large-scale GWASs using six common human diseases with sufficient GWAS data. We show that while PRS accuracy has grown rapidly for years, the improvement pace from recent GWAS has decreased substantially, suggesting that further increasing GWAS sample size may translate into very modest risk discrimination improvement. We next investigated the factors influencing the maximum achievable prediction using recently released whole genome-sequencing data from 125K UK Biobank participants, and state-of-the-art modeling of polygenic outcomes. Our analyses point toward increasing the variant coverage of PRS, using either more imputed variants or sequencing data, as a key component for future improvement in prediction accuracy.

genetics↗

bioRxiv2023

A resampling-based approach to share reference panels

For many genome-wide association studies, imputing genotypes from a haplotype reference panel is a necessary step. Over the past 15 years, reference panels have become larger and more diverse, leading to improvements in imputation accuracy. However, the latest generation of reference panels is subject to restrictions on data sharing due to concerns about privacy, limiting their usefulness for genotype imputation. In this context, we propose RESHAPE, a method that employs a recombination Poisson process on a reference panel to simulate the genomes of hypothetical descendants after multiple generations. This data transformation helps to protect against re-identification threats and preserves important data attributes, such as linkage disequilibrium (LD) patterns and, to some degree, Identity-By-Descent (IBD) sharing, allowing for genotype imputation. Our experiments on gold standard datasets show that simulated descendants up to eight generations can serve as reference panels without significantly reducing genotype imputation accuracy. We suggest that this specific type of data anonymization could be used to generate synthetic reference panels available under less restrictive data sharing policies.

bioinformatics↗

bioRxiv2023

SURFBAT: a surrogate family-based association test building on large imputation reference panels

Genotype-phenotype association tests are typically adjusted for population stratification using principal components that are estimated genome-wide. This lacks resolution when analysing populations with fine structure and/or individuals with fine levels of admixture. This can affect power and precision, and is a particularly relevant consideration when control individuals are recruited using geographic selection criteria. Such is the case in France where we have recently created reference panels of individuals anchored to different geographic regions. To make correct comparisons against case groups, who would likely be gathered from large urban areas, new methods are needed. We present SURFBAT (a SURrogate Family Based Association Test) which performs an approximation of the transmission-disequilibrium test. Our method hinges on the application of genotype imputation algorithms to match similar haplotypes between the case and control groups. This permits us to approximate local ancestry informed posterior probabilities of un-transmitted parental alleles of each case individual. SURFBAT provides an association test that is inherently robust to fine-scale population stratification and opens up the possibility of efficiently using large imputation reference panels as control groups for association testing. The method is suitable when the control panel spans the local ancestry spectrum of the case-group population and each control has similar paternal and maternal ancestries. This is the case for our reference panels where individuals have their four grand-parents born in the same geographic area. In contrast to other methods for association testing that incorporate local-ancestry inference, SURFBAT does not require a set of ancestry groups to be defined, nor for local ancestry to be explicitly estimated. We demonstrate the interest of our tool on simulated datasets created from the 1000 Genomes project and the FranceGenRef project, as well as on a real-data example for a group of case individuals affected by Brugada syndrome.

genetics↗

bioRxiv2022

Imputation of low-coverage sequencing data from 150,119 UK Biobank genomes

Recent work highlights the advantages of low-coverage whole genome sequencing (lcWGS), followed by genotype imputation, as a cost-effective genotyping technology for statistical and population genetics. The release of whole genome sequencing data for 150,119 UK Biobank (UKB) samples represents an unprecedented opportunity to impute lcWGS with high accuracy. However, despite recent progress1,2, current methods struggle to cope with the growing numbers of samples and markers in modern reference panels, resulting in unsustainable computational costs. For instance, the imputation cost for a single genome is 1.11{pound} using GLIMPSE v1.1.1 (GLIMPSE1) on the UKB research analysis platform (RAP) and rises to 242.8{pound} using QUILT v1.0.4. To overcome this computational burden, we introduce GLIMPSE v2.0.0 (GLIMPSE2), a major improvement of GLIMPSE, that scales sublinearly in both the number of samples and markers. GLIMPSE2 imputes a low-coverage genome from the UKB reference panel for only 0.08{pound} in compute cost while retaining high accuracy for both ancient and modern genomes, particularly at rare variants (MAF < 0.1%) and for very low-coverage samples (0.1x-0.5x).

bioinformatics↗

bioRxiv2022

Accurate rare variant phasing of whole-genome and whole-exome sequencing data in the UK Biobank

The UK Biobank performed whole-genome sequencing (WGS) and whole-exome sequencing (WES) across hundreds of thousands of individuals, allowing researchers to study the effects of both common and rare variants. Haplotype phasing distinguishes the two inherited copies of each chromosome into haplotypes and unlocks novel analyses at the haplotype level. In this work, we describe a new phasing method, SHAPEIT5, that accurately and rapidly phases large sequencing datasets and illustrates its key features on the UK Biobank WGS and WES data. First, we show that it phases rare variants with high accuracy. For instance, variants found in 1 sample out of 100,000 in the WES data are phased with accuracy above 95%. Second, we show that it can phase singletons, although with moderate accuracy, thereby making their inclusion in downstream analyses possible. Third, we show that the use of UK Biobank as a reference panel increases the accuracy of genotype imputation, an increase that is more pronounced when phased with SHAPEIT5 compared to other methods. Finally, we screen the phased WES data for loss-of-function (LoF) compound heterozygous (CH) events and identify 549 genes in which both gene copies are found knocked out. This list of genes complements current knowledge of gene essentiality in the human genome. We provide SHAPEIT5 in an open-source format, providing researchers with the means to leverage haplotype information in genetic studies.

genomics↗

bioRxiv2022

Widespread enhancer co-activity identified by multimodal single cell analysis

Non-coding regulatory elements such as enhancers are key in controlling the cell type-specificity and spatio-temporal expression of genes. To drive stable and precise gene transcription that is robust to genetic variation and environmental stress, genes are often targeted by multiple enhancers with redundant action. However, it is unknown whether enhancers targeting the same gene display simultaneous activity or whether some enhancer combinations are more often co-active than others. Here, we take advantage of the recent developments in single cell technology that permit assessing chromatin status (scATAC-seq) and gene expression (scRNA-seq) in the same single cells to link gene expression to the activity of multiple enhancers. Measuring activity patterns across 24,844 human lymphoblastoid single cells, we found that the majority of enhancers associated with the same gene display significant correlation in their chromatin profiles. For 6944 expressed genes associated with enhancers, we identified 89,885 significant enhancer-enhancer associations between nearby enhancers. We found that associated enhancers share similar transcription factor binding profiles and that gene essentiality is linked with higher enhancer co-activity. Our extensive enhancer co-activity maps can be used to pinpoint combinations of enhancers relevant in gene expression regulation and allow us to better predict the effect of genetic variation falling in non-coding regions.

genomics↗

bioRxiv2022

Genetic variation in correlated regulatory region of Immunity

Studying the interplay between genetic variation, epigenetic changes and regulation of gene expression in immune cells is important to understand the modification of cellular states in various conditions, including immune diseases. Here, we built cis maps of regulatory regions with coordinated activity - Cis Regulatory Domains (CRDs) - in neutrophils, monocytes and T cells. For this, we leveraged (i) whole-genome sequencing (WGS), (ii) chromatin immunoprecipitation sequencing (ChIP-seq), (iii) DNA methylation (450k arrays), and (iv) transcriptional profiles (RNA-seq) from the BLUEPRINT consortium, for up to 200 individuals. Our study uncovers 9287, 7666 and 5480 histone CRDs (hCRDs) and 6053, 6112, 5701 methyl CRDs (mCRDs) in monocytes, neutrophils and T-cells, respectively. We discovered 15294 hCRD-gene and 6185 mCRD-gene associations (5% FDR). Only 33% of hCRD-gene associations and 37% of mCRD-gene associations were shared between cell-types, revealing the dynamic nature of regulatory interactions and how similarly located regulatory regions modulate the activity of different genes on different cell types. We mapped Quantitative Trait Loci associated with CRD activity (CRD-QTLs) and found that 89% and 70% of these hCRDs and mCRDs are under genetic control highlighting the importance of genetic variation to study the coordination of cellular regulatory programs. We found CRD-QTLs to be enriched in celltype-specific transcription factor binding sites, such as SPI1, STAT3, RFX1, SOX4, ATF3 for neutrophils and monocytes and TCF4 and BCL11A for T-cells, in line with the Human protein Atlas. We integrated PCHi-C data, which showed that most significant associations discovered within gene-CRD associations and co-expressed genes associated with the same CRD, involving large genomic distances, tend to happen between genomic regions in close spatial proximity. Finally, we mapped trans regulatory associations between CRDs, which enabled the discovery of 207 trans-eQTLs across cell types. Overlapping our hits with trans eQTLs from eQTLGen Consortium meta-analysis in whole blood revealed 52 trans-eQTLs shared between the two studies. Overall, we show that mapping functional regulatory units using population genomics data allows discovering important mechanisms in the regulation of gene expression in immune cells and gain a greater understanding of cell-type specific regulatory mechanisms of immunity.

bioinformatics↗

bioRxiv2022

Imputation of ancient genomes

Due to postmortem DNA degradation, most ancient genomes sequenced to date have low depth of coverage, preventing the true underlying genotypes from being recovered. Genotype imputation has been put forward to improve genotyping accuracy for low-coverage genomes. However, it is unknown to what extent imputation of ancient genomes produces accurate genotypes and whether imputation introduces bias to downstream analyses. To address these questions, we downsampled 43 ancient genomes, 42 of which are high-coverage (above 10x) and three constitute a trio (mother, father and son), from different times and continents to simulate data with coverage in the range of 0.1x-2.0x and imputed these using state-of-the-art methods and reference panels. We assessed imputation accuracy across ancestries and depths of coverage. We found that ancient and modern DNA imputation accuracies were comparable. We imputed most of the 42 high-coverage genomes downsampled to 1x with low error rates (below 5%) and estimated higher error rates for African genomes, which are underrepresented in the reference panel. We used the ancient trio data to validate imputation and phasing results using an orthogonal approach based on Mendels rules of inheritance. This resulted in imputation and switch error rates of 1.9% and 2.0%, respectively, for 1x genomes. We further compared the results of downstream analyses between imputed and high-coverage genomes, notably principal component analysis (PCA), genetic clustering, and runs of homozygosity (ROH). For these three approaches, we observed similar results between imputed and high-coverage genomes using depths of coverage of at least 0.5x, except for African genomes, for which the decreased imputation accuracy impacted ROH estimates. Altogether, these results suggest that, for most populations and depths of coverage as low as 0.5x, imputation is a reliable method with potential to expand and improve ancient DNA studies.

genomics↗

bioRxiv2022

Accuracy of haplotype estimation and whole genome imputation affects complex trait analyses in complex biobanks

Sample recruitment for research consortia, hospitals, biobanks, and personal genomics companies span years, necessitating genotyping in batches, using different technologies. As marker content on genotyping arrays varies systematically, integrating such datasets is non-trivial and its impact on haplotype estimation (phasing) and whole genome imputation, necessary steps for complex trait analysis, remains under-evaluated. Using the iPSYCH consortium dataset, comprising 130,438 individuals, genotyped in two stages, on different arrays, we evaluated phasing and imputation performance across multiple phasing methods and data integration protocols. While phasing accuracy varied both by choice of method and data integration protocol, imputation accuracy varied mostly between data integration protocols. We demonstrate an attenuation in imputation accuracy within samples of non-European origin, highlighting challenges to studying complex traits in diverse populations. Finally, imputation errors can modestly bias association tests and reduce predictive utility of polygenic scores. This is the largest, most comprehensive comparison of data integration approaches in the context of a large psychiatric biobank.

bioinformatics↗

bioRxiv2021

Shared regulation and functional relevance of local gene co-expression revealed by single cell analysis

Most human genes are co-expressed with a nearby gene. Yet, previous studies only reported this extensive local gene co-expression using bulk RNA-seq. Here, we leverage single cell datasets in >85 individuals to identify gene co-expression across cells, unbiased by cell type heterogeneity and benefiting from the co-occurrence of transcription events in single cells. We discover thousands of co-expressed genes in two cell types and (i) compare single cell to bulk RNA-seq in identifying local gene co-expression, (ii) show that many co-expressed genes - but not the majority - are composed of functionally-related genes and (iii) provide evidence that these genes are transcribed synchronously and their co-expression is maintained up to the protein level. Finally, we identify gene-enhancer associations using multimodal single cell data, which reveal that >95% of co-expressed gene pairs share regulatory elements. Our in-depth view of local gene co-expression and regulatory element co-activity advances our understanding of the shared regulatory architecture between genes.

genomics↗

bioRxiv2021

Parent-of-origin effects in the UK Biobank

Identical genetic variations can have different phenotypic effects depending on their parent of origin (PofO). Yet, studies focussing on PofO effects have been largely limited in terms of sample size due to the need of parental genomes or known genealogies. Here, we used a novel probabilistic approach to infer PofO of individual alleles in the UK Biobank that does not require parental genomes nor prior knowledge of genealogy. Our model uses Identity-By-Descent (IBD) sharing with second- and third-degree relatives to assign alleles to parental groups and leverages chromosome X data in males to distinguish maternal from paternal groups. When combined with robust haplotype inference and haploid imputation, this allowed us to infer the PofO at 5.4 million variants genome-wide for 26,393 UK Biobank individuals. We used this large dataset to systematically screen 59 biomarkers and 38 anthropomorphic phenotypes for PofO effects and discovered 101 significant associations, demonstrating that this type of effects contributes to the genetics of complex traits. Notably, we retrieved well known PofO effects, such as the MEG3/DLK1 locus on platelet count, and we discovered many new ones at loci often unsuspected of being imprinted and, in some cases, previously thought to harbour additive associations.

genomics↗

bioRxiv2021

GCAT|Panel, a comprehensive structural variant haplotype map of the Iberian population from high-coverage whole-genome sequencing

The combined analysis of haplotype panels with phenotype clinical cohorts is a common approach to explore the genetic architecture of human diseases. However, genetic studies are mainly based on single nucleotide variants (SNVs) and small insertions and deletions (indels). Here, we contribute to fill this gap by generating a dense haplotype map focused on the identification, characterization and phasing of structural variants (SVs). By integrating multiple variant identification methods and Logistic Regression models, we present a catalogue of 35,431,441 variants, including 89,178 SVs ([≥]50bp), 30,325,064 SNVs and 5,017,199 indels, across 785 Illumina high coverage (30X) whole-genomes from the Iberian GCAT Cohort, containing 3.52M SNVs, 606,336 indels and 6,393 SVs in median per individual. The haplotype panel is able to impute up to 14,360,728 SNVs/indels and 23,179 SVs, showing a 2.7-fold increase for SVs compared with available genetic variation panels. The value of this panel for SVs analysis is shown through an imputed rare Alu element located in a new locus associated with mononeuritis of lower limb, a rare neuromuscular disease. This study represents the first deep characterization of genetic variation within the Iberian population and the first operational haplotype panel to systematically include the SVs into genome-wide genetic studies.

genomics↗