Multiple sclerosis is a chronic, immune-mediated demyelinating disease of the central nervous system, affecting nearly three million people worldwide. It is characterized by immune cell infiltration into the brain and spinal cord, destruction of the myelin sheaths that insulate nerve fibers, and progressive neurodegeneration. Decades of family and twin studies have established that genetics contributes substantially to disease risk, with heritability estimates ranging between 25 and 50 percent. The strongest single genetic signal lies within the human leukocyte antigen (HLA) region on chromosome 6, particularly the HLA-DRB1*15:01 allele, which confers a roughly threefold increase in risk. Beyond HLA, however, more than 200 non-HLA risk loci have been identified, each contributing only modest effects. Until now, nearly all of these discoveries have come from cohorts overwhelmingly composed of individuals of European ancestry, limiting both the precision and the portability of the resulting biological insights.

The new study tackled this limitation head-on through a multiancestry genome-wide association study (GWAS) of unprecedented scale. By pooling genetic and clinical data from tens of thousands of individuals with multiple sclerosis and comparable numbers of unaffected controls drawn from European, East Asian, African, Hispanic and Latin American, and other ancestry groups, the consortium was able to boost statistical power well beyond what any single-ancestry cohort could achieve. Combining ancestries in a single analysis increases the effective sample size, while trans-ancestry comparisons exploit differences in linkage disequilibrium patterns—the nonrandom association of variants across populations—to fine-map disease associations more precisely. When the same haplotype block is inherited differently across ancestries, the causal variant can be pinpointed by looking for the signal that remains consistent while surrounding markers shift.

This fine-mapping strategy allowed the researchers to narrow the credible sets of candidate causal variants at many loci, in some cases reducing lists of dozens of plausible candidates to just a handful. The team also identified novel risk loci that had gone undetected in European-only studies and demonstrated that some previously reported associations were population-specific, driven by alleles common in one ancestry but rare or absent in others. Several of these ancestry-specific signals were found in non-Europeans for the first time, underscoring the importance of diversity in genetic research and the risk of systematically overlooking disease biology in underrepresented populations. The consortium also developed and applied methods to transfer polygenic risk scores across ancestries, revealing both the promise and the current limitations of genetic risk prediction outside European populations.

Identifying associated regions, however, is only the first step. The vast majority of multiple sclerosis risk variants do not fall within protein-coding genes; instead, they cluster in regulatory regions of the genome—enhancers, promoters, and other noncoding elements that control when and where genes are switched on. To interpret these variants, the researchers assembled an extensive collection of multiomics datasets spanning the cell types most relevant to the disease. This included single-cell RNA sequencing to profile gene expression, single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq) to map open, actively regulatory chromatin, and epigenomic annotations such as histone modification marks that flag active enhancers and promoters. The datasets covered immune cells critical to the periphery of the disease—such as CD4-positive and CD8-positive T cells, B cells, monocytes, and natural killer cells—as well as central nervous system resident cells, including microglia, astrocytes, oligodendrocytes, and their precursors.

By integrating GWAS summary statistics with these cellular maps through statistical frameworks such as stratified linkage disequilibrium score regression and transcriptome-wide and chromatin-interaction-based colocalization analyses, the team could ask a deceptively simple question with profound implications: in which cells, and at which anatomical and developmental stages, does the genetic risk of multiple sclerosis actually operate? The answers were striking. Genetic risk was strongly enriched in regulatory elements active in immune cell populations, particularly those involved in T cell activation and adaptive immune responses, consistent with the established immunopathology of the disease. But the analyses also revealed significant enrichment in resident central nervous system cells—most notably microglia, the brain’s innate immune macrophages—supporting an emerging model in which both peripheral immune cells and CNS-resident cells contribute to disease initiation and progression.

Perhaps the most innovative aspect of the study was its spatiocellular dimension. Rather than treating tissues as homogeneous mixtures, the researchers leveraged spatially resolved transcriptomic data to map genetic risk onto defined anatomical regions of the brain and spinal cord. This approach revealed that risk variants were not uniformly distributed across the nervous system: certain regulatory programs, active in specific regions and specific cell populations within those regions, showed disproportionate enrichment of heritability. The findings suggest that the spatial context of gene regulation—where in the central nervous system a risk variant’s target gene is active—helps determine how genetic susceptibility manifests as focal inflammatory lesions, the histological hallmark of multiple sclerosis. This spatial perspective, enabled by recent advances in spatial transcriptomics, opens a new dimension in the interpretation of complex disease genetics that conventional bulk and even single-cell analyses cannot capture.

The multiomics integration also enabled the researchers to prioritize causal genes at risk loci, a task that has historically been one of the hardest problems in post-GWAS biology. At many loci, the nearest gene to a risk variant is not the gene through which the variant acts. Using chromatin contacts, expression quantitative trait locus (eQTL) data, and colocalization of association signals with gene expression, the team linked noncoding risk variants to their distal target genes across cell types. Several prioritized genes converged on biologically coherent pathways, including antigen presentation, cytokine signaling, T cell receptor signaling, and interferon response—pathways that are already the targets of existing therapies and that point toward new therapeutic opportunities. Notably, the integration of cross-ancestry data sharpened many of these gene mappings, because fine-mapping resolution improved when linkage disequilibrium patterns from multiple populations were combined.

The clinical implications of the work are considerable. More precise fine-mapping of causal variants improves the foundation for polygenic risk scores, which could eventually aid in identifying individuals at elevated risk before symptom onset, particularly given that early treatment of multiple sclerosis is associated with substantially better outcomes. The study’s cross-ancestry framework also represents a corrective to a long-standing inequity in human genetics: individuals of non-European ancestry have been markedly underrepresented in GWAS, which has limited the accuracy of genetic risk prediction and the generalizability of biological conclusions worldwide. By demonstrating that multiancestry designs yield novel loci and finer resolution even for well-studied diseases, the research provides a template that other consortia studying complex diseases—from type 1 diabetes to rheumatoid arthritis to systemic lupus erythematosus—can follow.

The study also deepens understanding of the immunology of multiple sclerosis at a moment when therapeutics are rapidly evolving. Modern disease-modifying treatments, including anti-CD20 B cell depletion, S1P receptor modulators, and high-efficacy induction therapies, have transformed the disease course for many patients, but none reliably halt progression, and progression independent of relapse activity remains a major unmet need. The identification of genetic risk operating within microglia and other CNS-resident cells offers a mechanistic bridge between the peripheral immune processes targeted by current drugs and the compartmentalized central nervous system inflammation thought to drive progressive disease. Genes and regulatory programs prioritized through the spatiocellular analyses may point to targets capable of modulating the resident immune environment of the brain—therapeutic territory that has so far been difficult to reach.

As with any genetic study, important caveats remain. Fine-mapped variants are candidates, not proof; functional validation in experimental systems will be needed to confirm the causal mechanisms at each locus. The multiomics atlases, while extensive, still incompletely capture the full cellular diversity of human immune and nervous tissue, particularly in disease-relevant states such as activated microglia within lesions or tissue-resident lymphocyte populations. And even with multiancestry data, sample sizes for some ancestry groups remain modest relative to European cohorts, meaning that further global expansion of genetic studies will be needed to complete the picture. The authors and the broader field regard this work as a foundation rather than an endpoint: a demonstration that when genetics is combined with cellular, epigenomic, and spatial context across ancestrally diverse populations, the biology buried within genome-wide association signals becomes dramatically clearer.

Taken together, the study marks a turning point in multiple sclerosis genetics. It moves the field from lists of associated genomic regions toward a mechanistic, four-dimensional view of disease risk—one that incorporates cell type, gene regulatory circuitry, and anatomical location, and that embraces the full breadth of human genetic diversity. For a disease that has confounded researchers for more than a century, that perspective may prove to be the key to translating three decades of genetic discovery into therapies that work for every patient, in every population, at every stage of disease.

Complex diseases—including immune disorders, metabolic conditions, neurological illnesses and many cancers—rarely arise from a single genetic alteration. Instead, they reflect the combined influence of numerous variants, each often contributing a small increase or decrease in risk. These variants can be distributed across the genome and may affect gene regulation rather than the structure of a protein. Genome-wide association studies, or GWAS, have been highly effective at detecting regions linked to disease, but the strongest statistical signal in a region is not necessarily the causal variant. Multiple variants may be inherited together, a phenomenon known as linkage disequilibrium, making it difficult to determine which alteration is biologically decisive.

The approach described in the study focuses attention on individual genomic loci—the defined regions surrounding disease-associated signals—rather than treating all associations as equivalent. Locus-specific analysis can help distinguish the genetic architecture of one region from another, recognizing that different loci may operate through entirely different mechanisms. One region might influence disease by changing the expression of a nearby gene, while another could affect a regulatory element active only in a particular cell type. By separating these local patterns, researchers can reduce the risk of applying a single broad interpretation to genetically diverse signals.

Prioritization is the second major component of the framework. Once variants and candidate genes have been identified within a disease-associated locus, they must be ranked according to the strength and biological relevance of the available evidence. This process may incorporate genetic association data, regulatory annotations, gene expression, chromatin activity, cellular context and known functional relationships. The goal is not simply to produce a longer list of possible genes, but to focus attention on the candidates most likely to explain the observed disease signal. In principle, this can help connect statistical genetics with experiments that test molecular function.

The study’s title points to a further objective: unveiling the mechanisms underlying genetic risk, rather than merely cataloguing risk markers. Mechanistic interpretation is essential because disease-associated variants frequently occur in noncoding DNA. These regions do not encode proteins, but they can contain promoters, enhancers and other regulatory sequences that control when and where genes are active. A variant in such a region may alter the binding of a transcription factor, modify chromatin accessibility or change the communication between a regulatory element and its target gene. Understanding these effects requires analysis at the level of tissues, cell types and genomic neighborhoods.

A locus-specific framework may also help explain why the same disease can emerge through multiple biological routes. Genetic risk is often heterogeneous: different patients may carry risk variants that converge on a common clinical outcome while acting through distinct pathways. Some variants may influence immune activation, others cellular metabolism, tissue repair or neuronal signaling. Stratifying signals by locus can expose these separate routes and reveal whether they converge on shared molecular processes. This distinction matters for drug discovery, because a therapy aimed at one mechanism may benefit only a genetically defined subgroup rather than every patient diagnosed with the same condition.

The practical significance of such prioritization extends beyond the interpretation of published GWAS results. Researchers can use ranked candidate genes and variants to select targets for laboratory validation, including gene-editing experiments, reporter assays, perturbation screens and studies in disease-relevant cells. The framework may also support the integration of genomic findings with transcriptomic and epigenomic datasets, allowing investigators to ask whether a risk variant changes gene activity in the tissue where disease begins. Such cross-layer analysis is increasingly important as scientists move from static DNA sequences toward dynamic models of gene regulation.

The work also highlights the importance of statistical caution. Association does not prove causation, and computational prioritization cannot replace experimental confirmation. A candidate gene may appear compelling because it is active in a relevant tissue or participates in a known pathway, yet those features alone do not demonstrate that it mediates genetic risk. Similarly, a regulatory variant may be correlated with disease because it is inherited alongside the true causal alteration. Robust interpretation therefore depends on combining multiple independent lines of evidence and accounting for uncertainty at every stage. The value of the proposed strategy lies in organizing that evidence around specific loci and biological hypotheses.

As genomic datasets become larger and more diverse, the need for interpretable frameworks is becoming more urgent. Many genetic studies have historically overrepresented people of European ancestry, limiting the generalizability of their findings and complicating the discovery of population-specific risk patterns. Locus-level analysis and prioritization could provide a structured way to compare signals across populations, tissues and disease subtypes, although the effectiveness of any framework will depend on the quality and diversity of the data supplied to it. By directing researchers toward the most plausible genetic mechanisms, the study offers a pathway from statistical association to testable biology—an essential step toward more precise disease classification, improved therapeutic targeting and a clearer understanding of why complex diseases develop.

Subject of Research: Genetic risk mechanisms underlying complex diseases

Article Title: Locus-specific stratification and prioritization unveil genetic risk mechanism underlying complex diseases

Article References: Zhang, J., Liu, Q., Zhu, Y. et al. Locus-specific stratification and prioritization unveil genetic risk mechanism underlying complex diseases. Nat Commun (2026). https://doi.org/10.1038/s41467-026-76649-3

Image Credits: AI Generated

DOI: 10.1038/s41467-026-76649-3

Keywords: complex diseases, genetic risk, locus-specific stratification, variant prioritization, genome-wide association studies, regulatory genomics, disease mechanisms

The complexity inherent to diseases influenced by gene networks presents formidable obstacles; the vast combinatorial possibilities of gene sets make traditional analytical methods insufficient. Unlike monogenic disorders, which stem from mutations in single genes, complex diseases arise from coordinated dysregulation among numerous genes, a phenomenon obscured by sheer data dimensionality and statistical limitations. To overcome this, researchers have turned to advanced machine learning methodologies, culminating in the creation of the Transcriptome-Wide conditional Variational auto-Encoder, or TWAVE. This generative model amplifies sparse gene expression datasets, enabling the identification of causal gene ensembles that drive pathological states.

TWAVE’s capacity to simulate both diseased and healthy cellular states through limited but high-quality gene expression data is revolutionary. Gene expression profiles offer dynamic insights far beyond static DNA sequences, capturing temporal and environmental influences on cellular behavior. The model thus pivots from mere genotype analysis towards an integrated genotype-phenotype framework. Instead of isolating individual gene contributions, TWAVE systematically uncovers collective gene effects that propagate complex traits, pinpointing key players whose combined activity shifts cellular states. This paradigm shift allows researchers to transcend the reductive simplicity of single-gene associations and capture the multigenic orchestration of disease.

The methodology capitalizes on a fusion of machine learning and optimization algorithms to enhance interpretability and causal inference. By training on clinical trial datasets with well-characterized expression states, TWAVE aligns gene expression perturbations with phenotypic manifestations. Moreover, experimental perturbation data, elucidating gene network responses to activation or suppression, refine the network inference, thus enhancing biological plausibility and predictive utility. This sophisticated interplay between data-driven modeling and experimental grounding highlights TWAVE’s potential in precision medicine.

Environmental modulation of gene activity further complicates the genetic landscape of complex traits. Traditional genomic studies often neglect these influences because DNA sequences remain constant regardless of environmental context. However, gene expression profiles fluctuate in response to external stimuli, offering a real-time depiction of cellular adaptation. TWAVE’s reliance on expression data elegantly integrates these environmental effects, providing a more holistic understanding of disease etiology. This dynamic integration signifies a critical leap forward in unraveling the genotype-to-phenotype enigma.

The implications of this approach extend beyond theoretical insights, promising tangible clinical applications. TWAVE has been validated across multiple complex diseases, demonstrating its superiority over conventional genome-wide association studies that often fail to detect subtle gene-gene interactions. Notably, it identifies gene sets that have previously eluded detection, underscoring the model’s sensitivity and robustness. Intriguingly, the research reveals interindividual heterogeneity in genetic drivers, suggesting that diseases traditionally classified under a single umbrella may, in reality, represent diverse molecular syndromes requiring personalized therapeutic strategies.

This personalized dimension is crucial in an era striving for precision medicine. Different patients manifesting phenotypically similar diseases may harbor distinct causal gene constellations, influenced by their unique genetic backgrounds, environmental exposures, and lifestyles. TWAVE’s ability to delineate these differences lays the groundwork for bespoke treatments, targeting patient-specific gene networks rather than one-size-fits-all interventions. Such precision could transform therapeutic efficacy and reduce adverse effects by tailoring interventions at the molecular level.

The significance of focusing on gene expression over gene sequence lies in its ethical and practical advantages. Because expression data sidesteps many privacy concerns associated with DNA sequencing, it facilitates broader data sharing and integration, accelerating research and discovery. Additionally, gene expression captures epigenetic and post-transcriptional regulatory influences, offering a richer, more functional perspective on genetic contributions to disease.

Senior author Adilson Motter, a physicist specializing in complex systems, likens the disease mechanism to an airplane crash requiring multiple concurrent failures. This analogy encapsulates the essence of polygenic traits: the convergence of multiple subtle genetic perturbations resulting in disease phenotypes. Motter’s vision is to disentangle these convoluted networks, bringing clarity to the chaos of biological complexity through TWAVE’s generative framework.

The model’s development reflects a multidisciplinary synergy, combining expertise in physics, computational biology, and genetics. Postdoctoral researcher Benjamin Kuznets-Speck, graduate student Buduka Ogonor, and research associate Thomas Wytock contributed extensively, integrating algorithmic innovation with biological insight. Their collective efforts underscore the necessity of collaborative research frameworks to tackle the most daunting challenges in biomedical science.

This pioneering research received support from esteemed institutions, including the National Cancer Institute, the National Science Foundation, and the Simons Foundation. The study’s forthcoming publication in the Proceedings of the National Academy of Sciences marks a milestone in computational biology and genomics, promising to catalyze further innovation in understanding and treating complex diseases.

In sum, Northwestern University’s TWAVE represents a monumental stride in biomedical research, harnessing generative AI to elucidate the complex genetic architectures of diseases long deemed inscrutable. By transcending traditional approaches and embracing the dynamic intricacies of gene expression, this tool opens new vistas for personalized medicine, promising a future where diagnostics and therapeutics are profoundly precise and tailored to the individual molecular landscapes that define human health and disease.

Subject of Research: Computational modeling of gene expression to identify causal gene sets underlying complex diseases

Article Title: Generative prediction of causal gene sets responsible for complex traits

News Publication Date: 9-Jun-2025

Web References:
DOI link

References:
Motter, A.E., Kuznets-Speck, B., Ogonor, B., & Wytock, T. (2025). Generative prediction of causal gene sets responsible for complex traits. Proceedings of the National Academy of Sciences. DOI: 10.1073/pnas.2415071122

Image Credits: Camila Felix

Keywords: Genotypes, Complex traits, Phenotypes, Gene expression, Genome dynamics, Medical genetics, Genetic disorders, Artificial intelligence, Generative AI