Biology subjects

Casale, F. P.

Publications and source records attributed to Casale, F. P..

17 recordsLinked to original sources

bioRxiv2026

Joint localization and testing reveals spatial context of differential expression

Multi-subject spatial transcriptomics is increasingly used to identify molecular differences across individuals and conditions while resolving where in tissue those differences occur. Yet subject-level differential expression (DE) analyses commonly require spatial regions to be specified in advance, separating localization from testing. Here we present SpaceDX, a framework that uses gene-specific attention-based spatial weighting to jointly test differential expression and localize the spatial expression patterns contributing to these associations. In simulations, SpaceDX outperforms cluster-based DE methods when differential expression does not align with predefined spatial boundaries. Across two multi-subject mouse brain datasets, SpaceDX detected 45.8% more DE genes than regional DESeq2 pseudobulk in a chronic stress model, while in an Alzheimer's disease model combining SpaceDX with DESeq2 increased discovery by 19.1% relative to DESeq2 alone. Across both datasets, the learned maps provided spatial context for condition-associated expression signals in addition to gene-level significance. In the chronic stress model, for example, oligodendrocyte- and myelin-associated genes showed shared localization to corresponding brain structures. Together, these results show that the spatial context of differential expression can be inferred jointly with gene-level associations, recovering signals obscured by regional aggregation without requiring the relevant tissue context to be defined in advance.

bioinformatics↗

bioRxiv2026

Benchmarking gene expression reconstruction from single-cell latent representations

Single-cell transcriptomics is typically modeled in low-dimensional latent representations that improve the signal-to-noise ratio of the data. Such representations underpin data integration, cell state discovery, and perturbation prediction, with applications ranging from large-scale organ atlases to latent trajectory modeling. Recent virtual cell approaches further leverage these representations to predict cellular responses as distributional shifts in latent space. Each of these applications ultimately requires faithful gene expression reconstruction from latent spaces for biological interpretation, enabling gene-level analysis of predicted perturbed or batch-corrected cells. Yet representation choice is typically treated as an implementation detail rather than a primary modeling decision, with no systematic evaluation of how well latent representations support gene expression reconstruction. Here, we introduce ReconEval, a benchmark for evaluating gene expression reconstruction from single-cell latent spaces. We benchmark two classes of latent representations: end-to-end trained models such as PCA, autoencoders, and variational autoencoders, and pretrained single-cell foundation model embeddings coupled to newly trained decoders. Reconstruction is evaluated both directly and after latent-space perturbation prediction. Across perturbational and observational datasets totaling over 100 million cells, our metric suite quantifies statistical fidelity; biological signal preservation, including differential expression, coexpression, cell-cycle structure, cytokine response and pathway activity; and perturbation-specific effects. We find that autoencoders achieve the strongest stand-alone reconstruction at low dimensionality, while variational regularization does not improve generalization in reconstruction. Frozen foundation model embeddings retain recoverable gene-level information, with reconstruction quality depending strongly on decoder architecture and pretraining objective. In latent perturbation modeling, high-dimensional PCA matches foundation model embeddings, while low-dimensional AE embeddings are optimal for flow-based generative models. Overall, reconstruction depends critically on the interplay between representation and downstream model, and simpler representations can outperform complex alternatives given appropriate capacity. Our benchmark establishes reconstruction as a critical evaluation axis for single-cell foundation models. We envision it improving the biological interpretability of latent-space modeling, a prerequisite for future virtual cell models to be validated by domain experts and grounded in biology.

bioinformatics↗

bioRxiv2026

Phenome-wide genetic framework to identify mechanisms of social effects

Phenotypes are shaped not only by an individuals genotype (direct genetic effects, DGE) and environment but also by the genetic composition of social partners through indirect genetic effects (IGE). Although IGE have been detected across many traits and species, their mechanisms remain largely unknown, particularly for physiological traits. Here we introduce a phenome-wide genetic framework that identifies proxy phenotypes for the heritable traits of social partners mediating IGE by estimating genetic correlations between IGE on focal phenotypes and DGE on measured traits. Applying this approach to two large, outbred mouse datasets comprising hundreds of behavioural and physiological phenotypes, we find that behavioural traits are neither more affected by IGE nor better proxies for the traits mediating them, challenging the prevailing behavioural-centric view of social effects. Instead, immune, metabolic and growth phenotypes are both affected by IGE and informative proxies for their underlying mechanisms, potentially reflecting the social transmission of gut microbes. Our framework provides a novel strategy to better understand the genetic basis of complex traits and uncover mechanisms of social effects.

genetics↗

bioRxiv2025

VariantFormer: A hierarchical transformer integrating DNA sequences with genetic variations and regulatory landscapes for personalized gene expression prediction

1Accurately predicting gene expression from DNA sequence remains a central challenge in human genetics. Current sequence-based models overlook natural genetic variation across individuals, while population-based models are restricted to variants observed within specific cohorts. Here, we present VariantFormer, a 1.2-billion-parameter transformer that predicts gene-level RNA abundance directly from personalized diploid genomes. Trained on 21,004 genome-transcriptome pairs from 2,330 donors, VariantFormer achieves state-of-the-art performance across both sequence- and population-based prediction tasks, while generalizing better to out-of-distribution contexts--including somatic mutation settings in cancer cell lines--and main-taining robustness across ancestries. Beyond expression prediction, VariantFormer improves eQTL effect size estimation compared to prior methods, with notable gains for lower-frequency and ancestry-specific variants. In applications to Alzheimers disease, VariantFormer gene embeddings prioritize likely causal genes and relevant tissue contexts, and in silico mutagenesis of known APOE alleles faithfully recovers known risk modifying effects. Together, these results establish VariantFormer as a scalable, diploid-aware framework for variant interpretation and personalized gene expression modeling across tissues and populations.

genomics↗

bioRxiv2025

Bayesian Aggregation of Multiple Annotations Enhances Rare Variant Association Testing

Gene-level rare variant association tests (RVATs) are essential for uncovering disease mechanisms and identifying potential drug targets. Advances in sequence-based machine learning have expanded the availability of variant pathogenicity scores, offering new opportunities to improve variant prioritization for RVATs. However, existing methods often rely on rigid models or analyze single annotations in isolation, limiting their ability to leverage these advances. To address this, we introduce BayesRVAT, a Bayesian framework for RVATs that models variant effects using multiple annotations. By specifying priors on variant effects and estimating gene-trait-specific burden scores through variational inference, BayesRVAT flexibly accommodates diverse genetic architectures. In simulations, BayesRVAT improved power while maintaining statistical calibration. When applied to real data from the UK Biobank, BayesRVAT detected 10.2% more associations with blood traits than the next-best method and uncovered novel gene-disease associations in analyses of eight disease traits, including a link between PRPH2 and retinal disease. These results highlight BayesRVATs potential to enhance rare variant discovery and improve gene-trait association studies.

genetics↗

bioRxiv2024

Tissue-specific apparent mtDNA heteroplasmy and its relationship with ageing and mtDNA gene expression

Heteroplasmy in the mitochondrial DNA (mtDNA) accumulates with age, but how much it occurs in tissues and how it affects function and physiology is not well-studied. Here we present a comprehensive tissue-specific map of mtDNA heteroplasmy and mtRNA modifications, and their relationships with age and mitochondrial gene expression. We propose a robust variant calling pipeline for mtDNA heteroplasmy and mtRNA modifications using bulk RNAseq data from 49 tissues in GTEx, and a calibrated phenotype association framework for both types of variants. We identify 109 associations between them and donor age, and 784 associations with tissue-specific mtDNA gene expression. Of these, 7 and 18 show cell-type specificity within tissue. In addition, we find 9 instances where these variations mediate donor age effects on mtDNA gene expression. Finally, we confirm previously identified relationships between mt-tRNA modifications and gene expression on their 5, but read- and gene-level investigations reveal previously unmodeled complexities.

genomics↗

bioRxiv2024

Functional gene embeddings improve rare variant polygenic risk scores

Rare variant association testing is a powerful strategy for identifying effector genes underlying common traits. However, its effectiveness is limited by the scarcity of high-impact rare allele carriers, posing challenges for sensitivity and robustness. Here, we introduce FuncRVP, a rare variant association framework addressing this issue by leveraging functional information across genes. FuncRVP models the effects of rare variants as a weighted sum of gene impairment scores, with weights regularized through a prior based on functional gene embeddings. Modeling 41 quantitative traits from unrelated UK Biobank participants showed that FuncRVP consistently outperformed linear regressions on significantly associated genes and did so more effectively for traits with higher burden heritability. The framework demonstrated versatility, yielding consistent improvements across diverse gene embeddings. Moreover, FuncRVP generated more robust gene effect estimates and yielded more gene discoveries, especially among genetically constrained genes. These findings demonstrate the value of integrating functional information in rare variant association studies and showcase FuncRVP as a promising tool for enhancing phenotype prediction and gene discovery.

genetics↗

bioRxiv2024

HistoGWAS: An AI-enabled Framework for Automated Genetic Analysis of Tissue Phenotypes in Histology Cohorts

Understanding how genetic variation affects tissue structure and function is crucial for deciphering disease mechanisms, yet comprehensive methods for genetic analysis of tissue histology are lacking. We address this gap with HistoGWAS, a framework integrating AI tools for representation learning and image generation with fast variance component models to enable scalable and interpretable genome-wide association studies of histological traits. HistoGWAS employs histology foundation models for automated trait characterization and generative models to visually interpret the genetic influences on these traits. Applied to eleven tissue types from the GTEx cohort, HistoGWAS identifies four genome-wide significant loci, which we linked to specific tissue histological and gene expression changes. A power analysis confirms the effectiveness of HistoGWAS in analyses of large-scale histological data, underscoring its potential to transform imaging genetic studies.

genetics↗

bioRxiv2023

Aberrant expression prediction across human tissues

Despite the frequent implication of aberrant gene expression in diseases, algorithms predicting aberrantly expressed genes of an individual are lacking. To address this need, we compiled an aberrant expression prediction benchmark covering 8.2 million rare variants from 633 individuals across 49 tissues. While not geared toward aberrant expression, the deleteriousness score CADD and the loss-of-function predictor LOFTEE showed mild predictive ability (1-1.6% average precision). Leveraging these and further variant annotations, we next trained AbExp, a model that yielded 12% average precision by combining in a tissue-specific fashion expression variability with variant effects on isoforms and on aberrant splicing. Integrating expression measurements from clinically accessible tissues led to another two-fold improvement. Furthermore, we show on UK Biobank blood traits that performing rare variant association testing using the continuous and tissue-specific AbExp variant scores instead of LOFTEE variant burden increases gene discovery sensitivity and enables improved phenotype predictions.

genetics↗

bioRxiv2023

Missing view completion for multi-timepoints omics data via representation disentanglement and temporal knowledge transfer

Longitudinal multi-view omics data offer unique insights into the temporal dynamics of individual-level physiology, which provides opportunities to advance personalized healthcare. However, the common occurrence of incomplete views makes extrapolation tasks difficult, and there is a lack of tailored methods for this critical issue. Here, we introduce LEOPARD, an innovative approach specifically designed to complete missing views in multi-timepoint omics data. By disentangling longitudinal omics data into content and temporal representations, LEOPARD transfers the temporal knowledge to the omics-specific content, thereby completing missing views. The effectiveness of LEOPARD is validated on three benchmark datasets constructed with data from the MGH COVID study and the KORA cohort, spanning periods from 3 days to 14 years. Compared to conventional imputation methods, such as missForest, PMM, GLMM, and cGAN, LEOPARD yields the most robust results across the benchmark datasets. LEOPARD-imputed data also achieve the highest agreement with observed data in our analyses for age-associated metabolites detection, estimated glomerular filtration rate-associated proteins identification, and chronic kidney disease prediction. Our work takes the first step toward a generalized treatment of missing views in longitudinal omics data, enabling comprehensive exploration of temporal dynamics and providing valuable insights into personalized healthcare.

bioinformatics↗

bioRxiv2022

An allelic series rare variant association test for candidate gene discovery

Allelic series are of candidate therapeutic interest due to the existence of a dose-response relationship between the functionality of a gene and the degree or severity of a phenotype. We define an allelic series as a gene in which increasingly deleterious mutations lead to increasingly large phenotypic effects, and develop a gene-based rare variant association test specifically targeted for the identification of allelic series. Building on the well-known burden and sequence kernel association (SKAT) tests, we specify a variety of association models, covering different genetic architectures, and integrate these into a COding-variant Allelic Series Test (COAST). Through extensive simulations, we confirm that COAST maintains the type I error and improves power when the pattern of coding-variant effect sizes increases monotonically with mutational severity. We applied COAST to identify allelic series for 4 circulating lipid traits and 5 cell count traits among 145,735 subjects with available whole exome sequencing data from the UK Biobank. Compared with optimal SKAT (SKAT-O), COAST identified 29% more Bonferroni significant associations with circulating lipid traits, on average, and 82% more with cell count traits. All of the gene-trait associations identified by COAST have corroborating evidence either from rare-variant associations in the full cohort (Genebass, N = 400K), or from common variant associations in the GWAS catalog. In addition to detecting many gene-trait associations present in Genebass using only a fraction (36.9%) of the sample, COAST detects associations, such as ANGPTL4 with triglycerides, that are absent from Genebass but which have clear common variant support.

genetics↗

bioRxiv2018

Genome-wide association study of social genetic effects on 170 phenotypes in laboratory mice

The phenotype of one individual can be affected not only by the individuals own genotypes (direct genetic effects, DGE) but also by genotypes of interacting partners (indirect genetic effects, IGE). IGE have been detected using polygenic models in multiple species, including laboratory mice and humans. However, the underlying mechanisms remain largely unknown. Genome-wide association studies of IGE (igeGWAS) can point to IGE genes, but have not yet been applied to non-familial IGE arising from "peers" and affecting biomedical phenotypes. In addition, the extent to which igeGWAS will identify loci not identified by dgeGWAS remains an open question. Finally, findings from igeGWAS have not been confirmed by experimental manipulation. We leveraged a dataset of 170 behavioural, physiological and morphological phenotypes measured in 1,812 genetically heterogeneous laboratory mice to study IGE arising between same-sex, adult, unrelated laboratory mice housed in the same cage. We developed methods for igeGWAS in this context and identified 24 significant IGE loci for 17 phenotypes (FDR < 10%). There was no overlap between IGE loci and DGE loci for the same phenotype, which was consistent with the moderate genetic correlations between DGE and IGE for the same phenotype estimated using polygenic models. Finally, we fine-mapped seven significant IGE loci to individual genes and confirmed, in an experiment with a knockout model, that Epha4 gives rise to IGE on stress-coping strategy and wound healing. Our results demonstrate the potential for igeGWAS to identify IGE genes and shed some light into the mechanisms of peer influence.

genetics↗

bioRxiv2018

A linear mixed model approach to study multivariate gene-environment interactions

Different environmental factors, including diet, physical activity, or external conditions can contribute to genotype-environment interactions (GxE). Although high-dimensional environmental data are increasingly available, and multiple environments have been implicated with GxE at the same loci, multi-environment tests for GxE are not established. Such joint analyses can increase power to detect GxE and improve the interpretation of these effects. Here, we propose the structured linear mixed model (StructLMM), a computationally efficient method to test for and characterize loci that interact with multiple environments. After validating our model using simulations, we apply StructLMM to body mass index in UK Biobank, where our method detects previously known and novel GxE signals. Finally, in an application to a large blood eQTL dataset, we demonstrate that StructLMM can be used to study interactions with hundreds of environmental variables.

genetics↗

bioRxiv2018

LiMMBo: a simple, scalable approach for linear mixed models in high-dimensional genetic association studies

Genome-wide association studies have helped to shed light on the genetic architecture of complex traits and diseases. Deep phenotyping of population cohorts is increasingly applied, where multi-to high-dimensional phenotypes are recorded in the individuals. Whilst these rich datasets provide important opportunities to analyse complex trait structures and pleiotropic effects at a genome-wide scale, existing statistical methods for joint genetic analyses are hampered by computational limitations posed by high-dimensional phenotypes. Consequently, such multivariate analyses are currently limited to a moderate number of traits. Here, we introduce a method that combines linear mixed models with bootstrapping (LiMMBo) to enable computationally efficient joint genetic analysis of high-dimensional phenotypes. Our method builds on linear mixed models, thereby providing robust control for population structure and other confounding factors, and the model scales to larger datasets with up to hundreds of phenotypes. We first validate LiMMBo using simulations, demonstrating consistent covariance estimates at greatly reduced computational cost compared to existing methods. We also find LiMMBo yields consistent power advantages compared to univariate modelling strategies, where the advantages of multivariate mapping increases substantially with the phenotype dimensionality. Finally, we applied LiMMBo to 41 yeast growth traits to map their genetic determinants, finding previously known and novel pleiotropic relationships in this high-dimensional phenotype space. LiMMBo is accessible as open source software (https://github.com/HannahVMeyer/limmbo).\n\nAuthor summaryIn multi-trait genetic association studies one is interested in detecting genetic variants that are associated with one or multiple traits. Genetic variants that influence two or more traits are referred to as pleiotropic. Multivariate linear mixed models have been successfully applied to detect pleiotropic effects, by jointly modelling association signals across traits. However, these models are currently limited to a moderate number of phenotypes as the number of model parameters grows steeply with the number of phenotypes, raising a computational burden. We developed LiMMBo, a new approach for the joint analysis of high-dimensional phenotypes. Our method reduces the number of effective model parameters by introducing an intermediate subsampling step. We validate this strategy using simulations, where we apply LiMMBo for the genetic analysis of hundreds of phenotypes, detecting pleiotropic effects for a wide range of simulated genetic architectures. Finally, to illustrate LiMMBo in practice, we apply the model to a study of growth traits in yeast, where we identify pleiotropic effects for traits with formerly known genetic effects as well as revealing previously unconnected traits.

genomics↗

bioRxiv2017

Interactions between genetic variation and cellular environment in skeletal muscle gene expression

From whole organisms to individual cells, responses to environmental conditions are influenced by genetic makeup, where the effect of genetic variation on a trait depends on the environmental context. RNA-sequencing quantifies gene expression as a molecular trait, and is capable of capturing both genetic and environmental effects. In this study, we explore opportunities of using allele-specific expression (ASE) to discover cis acting genotype-environment interactions (GxE) - genetic effects on gene expression that depend on an environmental condition. Treating 17 common, clinical traits as approximations of the cellular environment of 267 skeletal muscle biopsies, we identify 10 candidate interaction quantitative trait loci (iQTLs) across 6 traits (12 unique gene-environment trait pairs; 10% FDR per trait) including sex, systolic blood pressure, and low-density lipoprotein cholesterol. Although using ASE is in principle a promising approach to detect GxE effects, replication of such signals can be challenging as validation requires harmonization of environmental traits across cohorts and a sufficient sampling of heterozygotes for a transcribed SNP. Comprehensive discovery and replication will require large human transcriptome datasets, or the integration of multiple transcribed SNPs, coupled with standardized clinical phenotyping.

genetics↗

bioRxiv2016

Joint genetic analysis using variant sets reveals polygenic gene-context interactions

Joint genetic models for multiple traits have helped to enhance association analyses. Most existing multi-trait models have been designed to increase power for detecting associations, whereas the analysis of interactions has received considerably less attention. Here, we propose iSet, a method based on linear mixed models to test for interactions between sets of variants and environmental states or other contexts. Our model generalizes previous interaction tests and in particular provides a test for local differences in the genetic architecture between contexts. We first use simulations to validate iSet before applying the model to the analysis of genotype-environment interactions in an eQTL study. Our model retrieves a larger number of interactions than alternative methods and reveals that up to 20% of cases show context-specific configurations of causal variants. Finally, we apply iSet to test for sub-group specific genetic effects in human lipid levels in a large human cohort, where we identify a gene-sex interaction for C-reactive protein that is missed by alternative methods.\n\nAuthor summaryGenetic effects on phenotypes can depend on external contexts, including environment. Statistical tests for identifying such interactions are important to understand how individual genetic variants may act in different contexts. Interaction effects can either be studied using measurements of a given phenotype in different contexts, under the same genetic backgrounds, or by stratifying a population into subgroups. Here, we derive a method based on linear mixed models that can be applied to both of these designs. iSet enables testing for interactions between context and sets of variants, and accounts for polygenic effects. We validate our model using simulations, before applying it to the genetic analysis of gene expression studies and genome-wide association studies of human blood lipid levels. We find that modeling interactions with variant sets offers increased power, thereby uncovering interactions that cannot be detected by alternative methods.

genetics↗

bioRxiv2016

Genome-wide Analysis of Differential Transcriptional and Epigenetic Variability Across Human Immune Cell Types

BackgroundA healthy immune system requires immune cells that adapt rapidly to environmental challenges. This phenotypic plasticity can be mediated by transcriptional and epigenetic variability.\n\nResultsWe applied a novel analytical approach to measure and compare transcriptional and epigenetic variability genome-wide across CD14+CD16- monocytes, CD66b+CD16+ neutrophils, and CD4+CD45RA+ naive T cells, from the same 125 healthy individuals. We discovered substantially increased variability in neutrophils compared to monocytes and T cells. In neutrophils, genes with hypervariable expression were found to be implicated in key immune pathways and to associate with cellular properties and environmental exposure. We also observed increased sex-specific gene expression differences in neutrophils. Neutrophil-specific DNA methylation hypervariable sites were enriched at dynamic chromatin regions and active enhancers.\n\nConclusionsOur data highlight the importance of transcriptional and epigenetic variability for the neutrophils key role as the first responders to inflammatory stimuli. We provide a resource to enable further functional studies into the plasticity of immune cells, which can be accessed from: http://blueprint-dev.bioinfo.cnio.es/WP10/hypervariability.

molecular biology↗