Biology subjects

Speed, D.

Publications and source records attributed to Speed, D..

15 recordsLinked to original sources

bioRxiv2026

SNPic: SNP Topic Modeling for Interpretable Clustering of Complex phenotypes

Genome-wide association studies (GWAS) have cataloged thousands of disease-associated variants, yet a central challenge remains: decoding the shared, pleiotropic architecture that links complex phenotypes. Existing approaches, including dimensionality reduction methods and regression genetic models, either lack interpretability or rely on external linkage disequilibrium (LD) reference panels, limiting their ability to recover coherent biological mechanisms. Here we introduce the SNP topic model (SNPic), a generative probabilistic framework that reframes GWAS summary statistics as a structured corpus and models genetic architecture using principles from Natural Language Processing (NLP). By treating phenotypes as documents and genes or the whole corpus of traits as words, SNPic applies topic models, e.g. Latent Dirichlet Allocation (LDA), to infer latent "genetic topics", representing interpretable, overlapping biological modules that jointly explain complex traits. This formulation enables simultaneous reconstruction of trait relationships and identification of their underlying molecular drivers. SNPic integrates two complementary schemes: Sumstat-as-word for capturing global phenotypic structure and Gene-as-word for resolving mechanistic detail, within a unified modeling framework. To ensure robustness, we introduce a stability-optimized inference pipeline based on bootstrap resampling, allowing data-driven selection of topic number and filtering of stochastic signals. Across extensive simulations, SNPic consistently outperforms conventional dimensionality reduction methods in recovering latent structure under both linear and non-linear, highly overlapping genetic architectures. Applied to integrated FinnGen and UK Biobank datasets, SNPic identifies reproducible genetic topics corresponding to distinct biological programs, including HLA-mediated immune processes and transporter-driven metabolic regulation, with strong tissue-specific support. The framework further generalizes across species, organizing complex traits in maize, Arabidopsis thaliana, and cattle into biologically coherent modules. Together, these results establish SNPic as a scalable and interpretable framework that shifts GWAS analysis from association cataloging toward the construction of an interpretable knowledge graph representing the latent semantic architecture of the genome. By unifying statistical genetics with NLP, SNPic reframes GWAS analysis as a probabilistic language modeling task, enabling the systematic decoding of complex trait architectures and delivering a systemic graph of cross-phenotype relationships.

genetics↗

bioRxiv2022

Publicly Available Privacy-preserving Benchmarks for Polygenic Prediction

Recently, several new approaches for creating polygenic scores (PGS) have been developed and this trend shows no sign of abating. However, it has thus far been challenging to determine which approaches are superior, as different studies report seemingly conflicting benchmark results. This heterogeneity in benchmark results is in part due to different outcomes being used, but also due to differences in the genetic variants being used, data preprocessing, and other quality control steps. As a solution, a publicly available benchmark for polygenic prediction is presented here, which allows researchers to both train and test polygenic prediction methods using only summary-level information, thus preserving privacy. Using simulations and real data, we show that model performance can be estimated with accuracy, using only linkage disequilibrium (LD) information and genome-wide association summary statistics for target outcomes. Finally, we make this PGS benchmark - consisting of 8 outcomes, including somatic and psychiatric disorders - publicly available for researchers to download on our PGS benchmark platform (http://www.pgsbenchmark.org). We believe this benchmark can help establish a clear and unbiased standard for future polygenic score methods to compare against.

genetics↗

bioRxiv2022

Correction for participation bias in the UK Biobank reveals non-negligible impact on genetic associations and downstream analyses

While large-scale volunteer-based studies such as the UK Biobank (UKBB) have become the cornerstone of genetic epidemiology, the study participants are rarely representative of their target population. Here, we aim to evaluate the impact of non-random participation in the UKBB, and to pin down areas of research that are particularly susceptible to biases when using non-representative samples for genome-wide discovery. By comparing 14 harmonized characteristics of the UKBB participants to that of a representative sample, we derived a model for participation probability. We then conducted inverse probability weighted genome-wide association analyses (wGWA) on 19 UKBB traits. Comparing the output obtained from wGWA (Neffective=94,643 - 102,215) to standard GWA analyses (N=263,464 - 283,749), we assessed the impact of participation bias on three estimated quantities, namely 1) genotype-phenotype associations, 2) heritability and genetic correlation estimates and 3) exposure-outcome causal effect estimates obtained from Mendelian Randomization. Participation bias can lead to both overestimation (e.g., cancer, education) and underestimation (e.g., coffee intake, depression/anxiety) of SNP effects. Novel SNPs were identified in wGWA for 12 of the included traits, highlighting SNPs missed as a result of participation bias. While the impact of participation bias on heritability estimates was small (average change in h2: 1.5%, maximum: 5%), substantial distortions were present for genetic correlations (average absolute change in rg: 0.07, maximum: 0.31) and Mendelian Randomization estimates (average absolute change in standardized estimates: 0.04, maximum: 0.15), most markedly for socio-behavioural traits including education, smoking and BMI. Overall, the bias mainly affected the magnitude of effects, rather than direction. In contrast, genome-wide findings for more molecular/physical traits (e.g., LDL, SBP) exhibited less bias as a result of selective participation. Our results highlight that participation bias can distort genomic findings obtained in non-representative samples, and we propose a viable solution to reduce such bias. Moving forward, more efforts ensuring either sample representativeness or correcting for participation bias are paramount, especially when investigating the genetic underpinnings of behaviour, lifestyles and social outcomes.

genetics↗

bioRxiv2022

Parasite genotype is a major predictor of mortality from visceral leishmaniasis

Abstract/SummaryO_ST_ABSBackgroundC_ST_ABSVisceral leishmaniasis (VL) is a potentially fatal disease mainly caused by Leishmania infantum in South America and L. donovani in Asia and Africa. Disease outcomes have been associated with patient genotype, nutrition, age, sex, comorbidities, and co-infections. In this study, we examine the effects of parasite genetic variation on VL disease severity in Brazil. MethodsWe collected and sequenced the genomes of 109 L. infantum isolates from patients in northeast Brazil and retrieved matching patient clinical data from medical records, including mortality, sex, HIV co-infection and laboratory data (creatinine, haemoglobin, leukocyte and platelet counts). We identified genetic differences between parasite isolates, including single nucleotide polymorphisms (SNPs), small insertions/deletions (indels), and variations in genic, intergenic, and chromosome copy numbers (copy number variants, CNVs). To describe associations between the parasite genotypes and clinical outcomes, we applied quantitative genetics methods of heritability and genome-wide association studies (GWAS), treating clinical outcomes as traits that may be influenced by parasite genotype. FindingsMultiple aspects of the genetic analysis indicate that parasite genotype affects clinical outcomes. We estimate that parasite genotype explains 83% chance of mortality (narrow sense heritability, h2 = 0{middle dot}83{+/-}0{middle dot}17), and has a significant relationship with patient sex (h2 = 0{middle dot}60{+/-}0{middle dot}27). Impacts of parasite genotype on other clinical traits are lower (h2 [≤]0{middle dot}34). GWAS analysis identified multiple parasite genetic loci that were significantly associated with clinical outcomes; 17 CNVs that were significantly associated with mortality, two with creatinine and one with bacterial co-infection, jaundice and HIV co-infection; and two SNPs/indels and six CNVs that associate with age, jaundice, HIV and bacterial co-infections, creatinine, and/or bleeding sites. InterpretationParasite genotype is an important factor in VL disease severity in Brazil. Our analysis indicates that specific genetic differences between parasites act as virulence factors, enhancing risks of severe disease and mortality. More detailed understanding of these virulence factors could be exploited for novel therapies. Author SummaryMultiple factors contribute to the risk of mortality from visceral leishmaniasis (VL), including, patient genotype, comorbidities, and nutrition. Many of these factors will be influenced by socio-economic biases 1. Our work suggests that the virulence of the infecting parasite is an important risk factor for mortality. We pinpoint some specific genomic markers that are associated with mortality, which can lead to a greater understanding of the molecular mechanisms that cause severe VL disease, to genetic markers for virulent parasites and to the development of drug and vaccine therapies.

evolutionary biology↗

bioRxiv2021

Genome-wide association, prediction and heritability in bacteria

Advances in whole-genome genotyping and sequencing have allowed genome-wide analyses of association, prediction and heritability in many organisms. However, the application of such analyses to bacteria is still in its infancy, being limited by difficulties including the plasticity of bacterial genomes and their strong population structure. Here we propose, and validate using simulations, a suite of genome-wide analyses for bacteria. We combine methods from human genetics and previous bacterial studies, including linear mixed models, elastic net and LD-score regression, and introduce innovations such as frequency-based allele coding, testing for both insertion/deletion and nucleotide effects and partitioning heritability by genome region. We then analyse three phenotypes of a major human pathogen Streptococcus pneumoniae, including the first analyses of minimum inhibitory concentrations (MIC) for each of two antibiotics, penicillin and ceftriaxone. We show that these are highly heritable leading to high prediction accuracy, which is explained by many genetic associations identified under good control of population structure effects. In the case of ceftriaxone MIC, these results are surprising because none of the isolates was resistant according to the inhibition zone diameter threshold. We estimate that just over half of the heritability of penicillin MIC is explained by a known drug-resistance region, which also contributes around a quarter of the heritability of ceftriaxone MIC. For the within-host survival phenotype carriage duration, no reliable associations were found but we observed moderate heritability and prediction accuracy, indicating a polygenic trait. While generating important new results for S. pneumoniae, we have critically assessed existing methods and introduced innovations that will be useful for future large-scale population genomics studies to help decipher the genetic architecture of bacterial traits. Author summaryGenome-wide association, prediction and heritability analyses in bacteria are beginning to help unravel the genetic underpinnings of traits such as antimicrobial resistance, virulence, within-host survival and transmissibility. Progress to date is limited by challenges including the effects of strong population structure and variable recombination, and the many gaps in sequence alignments including the absence of entire genes in many isolates. More work is required to critically asses and develop methods for bacterial genomics. We address this task here, using a range of existing methods from bacterial and human genetics, such as linear mixed models, elastic net and LD-score regression. Using simulations, we first validate and then adapt these methods to introduce new analyses, including separate assessment of gap and nucleotide effects, a new allele coding for association analyses and a method to partition heritability into genome regions. We analyse within-host survival and two antimicrobial response traits of Streptococcus pneumoniae, identifying many novel associations while demonstrating good control of population structure and accurate prediction. We present both new results for an important pathogen and methodological advances that will be useful in guiding future studies in bacterial population genomics.

genomics↗

bioRxiv2020

Signatures Of TSPAN8 Variants Associated With Human Metabolic Regulation And Diseases

Here, with the example of common copy number variation (CNV) in the TSPAN8 gene, we present an important piece of work in the field of CNV detection, CNV association with complex human traits such as 1H NMR metabolomic phenotypes and an example of functional characterization of CNVs among human induced pluripotent stem cells (HipSci). We report TSPAN8 exon 11 as a new locus associated with metabolomic regulation and show that its biology is associated with several metabolic diseases such as type 2 diabetes (T2D), obesity and cancer. Our results further demonstrate the power of multivariate association models over univariate methods and define new metabolomic signatures for several new genomic loci, which can act as a catalyst for new diagnostics and therapeutic approaches.

genomics↗

bioRxiv2020

Exploring the variance in complex traits captured by DNA methylation assays

Following years of epigenome-wide association studies (EWAS), traits analysed to date tend to yield few associations. Reinforcing this observation, we conducted EWAS on 400 traits and 16 yielded at least one association at the conventional significance threshold (P<1x10-7). To investigate why EWAS yield is low, we formally estimated the proportion of phenotypic variation captured by 421,693 blood derived DNA methylation markers (h2EWAS) across all 400 traits. The mean h2EWAS was zero, with evidence for regular cigarette smoking exhibiting the largest association with all markers (h2EWAS=0.42) and the only one surpassing a false discovery rate < 0.1. Though underpowered to determine the h2EWAS value for any one trait, h2EWAS was predictive of the number of EWAS hits across the traits analysed (AUC=0.7). Modelling the contributions of the methylome on a per-site versus a per-region basis gave varied h2EWAS estimates (r=0.47) but neither approach obtained substantially higher model fits across all traits. Our analysis indicates that most complex traits do not heavily associate with markers commonly measured in EWAS within blood. However, it is likely DNA methylation does capture variation in some traits and h2EWAS may be a reasonable way to prioritise traits that are likely to yield associations.

genetics↗

bioRxiv2020

Improved genetic prediction of complex traits from individual-level data or summary statistics

Most existing tools for constructing genetic prediction models begin with the assumption that all genetic variants contribute equally towards the phenotype. However, this represents a suboptimal model for how heritability is distributed across the genome. Therefore, we develop prediction tools that allow the user to specify the heritability model. We compare individual-level data prediction tools using 14 UK Biobank phenotypes; our new tool LDAK-Bolt-Predict outperforms the existing tools Lasso, BLUP, Bolt-LMM and BayesR for all 14 phenotypes. We compare summary statistic prediction tools using 225 UK Biobank phenotypes; our new tool LDAK-BayesR-SS outperforms the existing tools lassosum, sBLUP, LDpred and SBayesR for 223 of the 225 phenotypes. The increase in prediction accuracy from improving the heritability model tends to be substantial. For example, when using LDAK-Bolt-Predict, the proportion of phenotypic variance explained increased by on average 14% (range 5-29%), equivalent to increasing the sample size by a quarter.

genetics↗

bioRxiv2019

Evaluating and improving heritability models using summary statistics

There is currently much debate regarding the best way to model how heritability varies across the genome. The authors of GCTA recommend the GCTA-LDMS-I Model, the authors of LD Score Regression recommend the Baseline LD Model, while we have instead recommended the LDAK Model. Here we provide a statistical framework for assessing heritability models using summary statistics from genome-wide association studies. Using data from studies of 31 complex human traits (average sample size 136,000), we show that the Baseline LD Model is the most realistic of the existing heritability models, but that it can be improved by incorporating features from the LDAK Model. Our framework also provides a method for estimating the selection-related parameter from summary statistics. We find strong evidence (P<1e-6) of negative genome-wide selection for traits including height, systolic blood pressure and college education, and that the impact of selection is stronger inside functional categories such as coding SNPs and promoter regions.

genetics↗

bioRxiv2019

Estimating narrow-sense heritability from genome-wide data in admixed populations

Finding an efficient framework for estimating total narrow-sense heritability in admixed populations remains an open question. In this work, we used extensive simulations to evaluate existing linear mixed model frameworks in estimating total narrow-sense heritability in two population-based cohorts from Greenland and compared the results to data from unadmixed individuals from Denmark. When our analysis focused on Greenlandic sib pairs, the model with two relationship matrices, one capturing identity by descent and one capturing identity by state, returned heritability estimates close to the true simulated value, while using each of the two matrices alone led to downward biases. When phenotypes correlated with ancestry, heritability estimates were inflated. Based on these observations, we propose a post-estimation PCA-based adjustment that recovers successfully the true simulated heritability. We use this knowledge to estimate the heritability of ten quantitative traits from the two Greenlandic cohorts and report differences such as lower heritability for height in Greenlanders compared to Europeans. In conclusion, narrow-sense heritability in admixed populations is best estimated using a mixture of genetic relationship matrices on individuals with at least one first-degree relative included in the sample.

genetics↗

bioRxiv2019

Investigating the Association Between Body Fat and Depression via Mendelian Randomization

Obesity and depression are major public health concerns that are both associated with substantial morbidity and mortality. There is a considerable body of literature linking obesity to the development of depression. Recent studies using Mendelian Randomization indicate that this relationship is causal. Most studies of the obesity-depression association have used body mass index as a measure of obesity. Body mass index is defined as weight (measured in kilograms) divided by the square of height (meters) and therefore does not distinguish between the contributions of fat and non-fat to body weight. To better understand the obesity-depression association, we conduct a Mendelian Randomization study of the relationship between fat mass, non-fat mass, height, and depression, using genome-wide association study results from the UK Biobank (n=332,000) and the Psychiatric Genomics Consortium (n=480,000). Our findings suggest that both fat mass and height (short stature) are causal risk factors for depression, while non-fat mass is not. These results represent important new knowledge on the role of anthropometric measures in the etiology of depression. They also suggest that reducing fat mass will decrease the risk of depression, which lends further support to public health measures aimed at reducing the obesity epidemic.

genetics↗

bioRxiv2019

Summary statistic analyses do not correct confounding bias

LD SCore regression (LDSC) has become a popular approach to estimate confounding bias, heritability and genetic correlation using only genome wide association study (GWAS) test statistics. SumHer is a newly-introduced alternative with similar aims. We show using theory and simulations that both approaches fail to adequately account for confounding bias, even when the assumed heritability model is correct. Consequently, these methods may estimate heritability poorly if there was inadequate adjustment for confounding in the original GWAS analysis. We also show that choice of summary statistic for use in LDSC or SumHer can have a large impact on resulting inferences. Further, covariate adjustments in the original GWAS can alter the target of heritability estimation, which can be problematic when LDSC or SumHer is applied to test statistics from a meta-analysis of GWAS with different covariate adjustments.

genetics↗

bioRxiv2018

Does Neuroticism Cause Depression? A Mendelian Randomization Study

Neuroticism is a heritable personality trait, characterized by negative emotions such as worrying, feelings of guilt, loneliness and being easily hurt. Increased levels of neuroticism are associated with poor mental health - development of depression in particular - but it remains uncertain whether this association represents a causal effect [1]. Here, we use Mendelian randomization (MR) to investigate whether neuroticism is a causal risk factor for development of depression.\n\nMR is an analytic approach to assess the causality of an observed association between a risk factor and a clinically relevant outcome [2]. MR is particularly useful in situations where randomized controlled trials are not possible and observational studies provide biased associations because of confounding or reverse causality. MR uses genome-wide association study (GWAS) data to identify instrumental variables (single nucleotide polymorphisms; SNPs) for a risk f ...

genetics↗

bioRxiv2018

Better estimation of SNP heritability from summary statistics provides a new understanding of the genetic architecture of complex traits

LD Score Regression (LDSC) has been widely applied to the results of genome-wide association studies. However, its estimates of SNP heritability are derived from an unrealistic model in which each SNP is expected to contribute equal heritability. As a consequence, LDSC tends to over-estimate confounding bias, under-estimate the total phenotypic variation explained by SNPs, and provide misleading estimates of the heritability enrichment of SNP categories. Therefore, we present SumHer, software for estimating SNP heritability from summary statistics using more realistic heritability models. After demonstrating its superiority over LDSC, we apply SumHer to the results of 24 large-scale association studies (average sample size 121 000). First we show that these studies have tended to substantially over-correct for confounding, and as a result the number of genome-wide significant loci has under-reported by about 20%. Next we estimate enrichment for 24 categories of SNPs defined by functional annotations. A previous study using LDSC reported that conserved regions were 13-fold enriched, and found a further twelve categories with above 2-fold enrichment. By contrast, our analysis using SumHer finds that conserved regions are only 1.6-fold (SD 0.06) enriched, and that no category has enrichment above 1.7-fold. SumHer provides an improved understanding of the genetic architecture of complex traits, which enables more efficient analysis of future genetic data.

genetics↗

bioRxiv2018

Exposing flaws in S-LDSC; reply to Gazal et al.

In our recent publication,1 we examined the two heritability models most widely used when estimating SNP heritability: the GCTA Model, which is used by the software GCTA2 and upon which LD Score regression (LDSC) is based,3 and the LDAK Model, which is used by our software LDAK.4 First we demonstrated the importance of choosing an appropriate heritability model, by showing that estimates of SNP heritability can be highly sensitive to which model is assumed. Then we empirically tested the GCTA and LDAK Models on GWAS data for a wide variety of complex traits. We found that the LDAK Model fits real data both significantly and substantially better than the GCTA Model, indicating that LDAK estimates more accurately describe the genetic architecture of complex traits than those from GCTA or LDSC.\n\nSome of our most striking results were our revised estimates of functional enrichments (the heritability enrichments of SNP categories defined by functional annotations). In general, estimates from LDAK were substantially more modest than previous estimates based on the GCTA Model. For example, we estimated that DNase I hypersensitive sites (DHS) were 1.4-fold (SD 0.1) enriched, whereas a study using GCTA had found they were 5.1-fold (SD 0.5) enriched,5 and we estimated that conserved SNPs were 1.3-fold (SD 0.3) enriched, whereas a study using S-LDSC (stratified LDSC) had found they were 13.3-fold (SD 1.5) enriched.6\n\nIn their correspondence, Gazal et al. dispute our findings. They assert that the heritability model assumed by LDSC is more realistic than the LDAK Model, and that estimates of enrichment from S-LDSC7 are more accurate than those from LDAK. Here, we explain why their justification for preferring the model used by LDSC is incorrect, and provide a simple demonstration that S-LDSC produces unreliable estimates of enrichment.

genetics↗