Biology subjects

Hu, X.

Publications and source records attributed to Hu, X..

At least 55 records · Page 3Linked to original sources

bioRxiv2026

Combination of engineered cell type-specific promoters and a high-efficiency AAV capsid restores hearing in adult DFNB1 mice model with demonstrated safety in nonhuman primate

A major challenge in gene therapy for GJB2-related hearing loss (DFNB1)--the most common form of hereditary deafness--is achieving efficient and precise connexin 26 delivery. Herein, we engineered two cell type-specific promoters (GJB2-1 and WFS1-2274) and developed an AAV capsid, AAV-MAS012, with enhanced transduction efficiency in mature cochlear cells. Our AAV-mediated gene therapy systems restored hearing of low-to-mid-frequencies in newborn Gjb2 cKO mice to wild-type levels and maintained for 45 weeks. Additionally, our therapeutic systems restored low-to-mid-frequencies hearing function to wild-type levels in adult Gjb2 cKO mice. A humanized version of the therapy, AAV-MAS012-WFS1-2274-hGJB2, rescued hearing function in two distinct Gjb2-deficient mouse models, and demonstrated a favorable safety profile in nonhuman primates. This study represents the first successful hearing restoration in adult Gjb2-deficient mice. The significant therapeutic efficacy of the humanized gene therapy system shows great potential for clinical translation in DFNB1 patients.

genetics↗

bioRxiv2026

Multiscale transcriptomic organization of the human brain with DigitalBrain

The human brain varies across anatomical regions, cell types, development, aging and disease states, yet existing single-cell transcriptomic resources remain fragmented and difficult to integrate into a unified biological model. Here we present DigitalBrain, a human brain-specific atlas and foundation-model framework for organizing diverse and fragmented human brain transcriptomic data across scales. We first built DigitalBrain-Atlas, a harmonized whole-brain single-cell resource comprising 16.35 million transcriptomes from 2,143 donors across 165 brain regions, spanning the human lifespan and multiple neurological and clinical conditions. We then developed DigitalBrain-M1, a Transformer-based model that jointly encodes gene identity and expression magnitude to learn a shared embedding space for cells and genes. Across held-out datasets, DigitalBrain supported robust single-cell integration, clustering and cell-type annotation while preserving major biological structure and reducing technical fragmentation. Beyond these benchmarks, the learned embeddings revealed emergent large-scale hierarchical organization of the human brain, linking anatomically distinct regions into higher-order patterns consistent with known functional systems. Applied to human hippocampal aging, DigitalBrain identified cell-type-specific aging sensitive gene sets, highlighted dentate gyrus granule cells as a particularly age-sensitive population, and discovered selective reorganization of gene programs related to synaptic transmission, postsynaptic structure, membrane excitability and axon guidance during aging. Cross-dataset convergence was strongest at the level of functional modules and recurrent aging sensitive genes. Together, these results demonstrate that DigitalBrain is a brain-specific framework for mapping human brain organization across scales, and as an early step towards a complete virtual organ for the human brain.

bioinformatics↗

bioRxiv2026

Decoding Single-Cell Omics of Perturbation Responses Using DeSCOPE

Deciphering cellular responses to genetic perturbations is fundamental to modeling gene regulatory networks and understanding mechanisms that change cellular phenotypes. However, current computational approaches often fail to outperform simple baseline models, highlighting a critical bottleneck in their generalizability and robustness. Here, we present DeSCOPE, a lightweight conditional variational autoencoder framework for predicting genetic perturbation responses spanning transcriptomic, epigenomic, and broader multi-modal landscapes. We systematically benchmarked DeSCOPE across diverse datasets under two challenging out-of-distribution settings: unseen genes and unseen cell types. DeSCOPE uniquely surpasses simple baselines in the unseen gene scenario, and achieves substantially improved performance for unseen cell types while requiring fine-tuning with far fewer perturbed genes. Finally, DeSCOPE demonstrates superior performance in predicting combinatorial multi-gene perturbations. Overall, DeSCOPE serves as a versatile multi-modal virtual cell model that can effectively guide the design of therapeutic targets that change cellular phenotypes. DeSCOPE is available at https://github.com/wanglabtongji/DeSCOPE.

bioinformatics↗

bioRxiv2026

Retinoic acid resolves the conflict between X-chromosome inactivation and pluripotency program in female cleavage-stage embryos

The tight coupling of X-chromosome inactivation (XCI) and pluripotency is a paradigm of holistic developmental regulation, and their inverse correlation enables XCI establishment during implantation. In contrast, the co-establishment of imprinted XCI (iXCI) and pluripotency program in female preimplantation embryos challenges the holistic pattern, but how embryos coordinate the two remains unknown. Here, we find that Nanog during the cleavage stage strongly represses the co-upregulated Rnf12-Xist axis, the triggering signaling of iXCI, creating a conflict that threatens iXCI. We identify retinoic acid (RA) as the coordinator that resolves the conflict and safeguards iXCI. RA activates its receptor RARG to evict NANOG from Rnf12 5region, avoiding NANOGs threat. In vivo, either maternal vitamin A deficiency or embryonic RA synthesis defects results in impaired iXCI and female-biased embryonic lethality. Thus, we highlight RA as a master regulator to holistically balance iXCI and pluripotency during cleavage stages, thereby ensuring proper female embryogenesis.

developmental biology↗

bioRxiv2026

Curvature-mediated prewetting organize mitochondrial nucleoid

The emergence of spatiotemporal order in diffusive cellular environments requires physical mechanisms to overcome the entropic drive toward disorder. While the organizing role of chemical signaling is well characterized, how complex membrane curvature regulate soluble molecular organization remains a fundamental open question. Here, we establish membrane curvature as a thermodynamic control parameter for the condensation of soluble proteins through prewetting. Using the mitochondrial transcription factor TFAM, we demonstrate that intrinsic membrane curvature locally drives a prewetting transitiona surface-mediated phase separation distinct from bulk condensation. By combining in vitro reconstitution, live-cell super-resolution imaging, cryo-electron tomography, and thermodynamic theory, we show that negatively high-curvature regions lower the nucleation energy barrier of Tfam, driving localized protein condensation at physiological concentrations. These results reveal that membrane curvature does not merely scaffold cellular machinery but manipulates the local free energy landscape, establishing a geometric principle for the spatial control of biological processes.

biophysics↗

bioRxiv2026

Vitamin D3 attenuates nitrogen mustard-induced dermal toxicity by enhancing microbial butyrate production via the intestinal VDR-α-defensin signaling pathway

Nitrogen mustard (NM)-caused severe cutaneous damage lacks effective targeted therapies. Vitamin D3 (VD3) shows promise as a therapy for NM-induced dermal toxicity; however, the underlying mechanisms remain elusive. Herein, we initially confirmed that NM induced gut flora dysbiosis, characterized by a decrease of Akkermansia muciniphila (AKK) abundance, thereby leading to butyrate reduction. Antibiotics (ABX) significantly promoted NM-induced skin injury, whereas fecal microbiota transplantation of the controls feces (HC-FMT) or AKK administration attenuated NM-induced dermal toxicity. HC-FMT or AKK significantly increased butyrate levels in feces and serum of NM-treated mice. Butyrate notably attenuated ABX-caused acceleration of NM-induced skin injury. Meanwhile, NM markedly decreased the expression of -defensins, MMP7, and VDR. NM failed to further decrease AKK abundance and BA contents in intestinal MMP7-deficient mice, which was abolished by human alpha defensin 5 (HD5) overexpression. And intestinal MMP7 deficiency enhanced NM-caused skin injury, which was markedly attenuated by HD5 overexpression, AKK transplantation, or BA supplementation. Moreover, NM also failed to further reduce MMP7 and -defensin expression, AKK abundance, and butyrate levels in intestinal VDR-silenced mice. Finally, VD3 remodeled the gut microbiome particularly enriching AKK, increased butyrate contents and promoted the expression of -defensins, MMP7, and VDR, thereby attenuating NM-induced skin damage. The protective effect of VD3 against NM-caused dermal toxicity was abolished by either ABX or intestinal-specific knockdown of MMP7 or VDR in mice; however, this impairment was reversed by butyrate or AKK. In conclusion, VD3 attenuated NM-caused dermal toxicity by promoting BA production via remodeling the gut microbiota, and this effect was partially mediated by the intestinal VDR--defensin signaling pathway. These highlight that targeting the gut flora or supplementing with BA could be potential therapies for NM-induced dermal toxicity.

molecular biology↗

bioRxiv2026

A Multi-Dataset Transcriptomic Analysis Unravels Core Mechanisms Involving Vitamin D Metabolism and Inflammatory Pathways for Frailty Diagnosis.

Frailty is a prevalent geriatric syndrome, and the shortage of objective biomarkers restricts its early diagnosis and intervention. This study aimed to identify robust molecular signatures and diagnostic markers for frailty using bioinformatics analyses of multiple independent datasets. Two transcriptome datasets (GSE144304, n=80; GSE287726, n=70) were obtained from the GEO database. We performed differential gene expression analysis, GO, KEGG and GSEA enrichment, and machine learning (70% training / 30% validation) to screen and validate core biomarkers. Numerous shared differentially expressed genes were identified. Vitamin D metabolism, ABC transporter, and inflammatory/immune pathways were consistently enriched and confirmed by GSEA. Machine learning models based on these signatures showed favorable diagnostic performance. Our study demonstrates that vitamin D metabolic disorders and chronic inflammation are core molecular features of frailty. The identified biomarkers provide new strategies for basic research, early clinical diagnosis, and therapeutic target development for frailty.

bioinformatics↗

bioRxiv2026

Exploration of Orally Disintegrating Tablet for Sublingual Vaccination against Mucosal Bacterial Infection

Oral mucosal bacterial infections impose a substantial global disease burden, yet current clinical management typically reduces microbial load only transiently and rarely establishes durable protection at the oral surface. Analysis of 200 patients with periodontitis revealed that elevated levels of pathogen-specific salivary secretory Immunoglobulin A (sIgA) were strongly associated with reduced bacterial burden and improved clinical periodontal outcomes, identifying sIgA as a key determinant of effective oral protection. Guided by this observation, we developed a sublingual, orally disintegrating tablet vaccine (Capot) that incorporates bacterial extracellular vesicles providing a comprehensive repertoire of native antigens and multiple pathogen-associated molecular patterns, encapsulated within a calcium phosphate nanoshell to enable safe transmucosal delivery to submandibular lymph nodes. The rapidly disintegrating tablet format minimizes inadvertent swallowing and enhances local mucosal bioavailability. In mice and non-human primates, Capot induced robust and long-lasting salivary sIgA responses without overt oral mucosal or gastrointestinal inflammation and conferred strong protection against primary, recurrent, and antibiotic-resistant periodontitis. Together, these findings establish sublingual tablet vaccination as a practical strategy for selectively engaging oral mucosal immunity and preventing chronic bacterial diseases at oral mucosal surfaces.

bioengineering↗

bioRxiv2026

Indazolone-Based Molecular Glue Degraders as a Tunable Platform for Reprogramming Cereblon Substrate Specificity

Molecular glue degraders (MGDs) represent a transformative modality in drug discovery, with the cereblon (CRBN)-MGD axis offering a strategic gateway to systematically address the historically undruggable proteome. However, the limited chemical diversity of existing CRBN-mediated MGDs has constrained the full exploration of the degradable landscape. Here, we introduce an innovative indazolone-based platform that extends beyond canonical isoindolinone-glutarimide scaffolds, significantly expanding the accessible chemical space for CRBN modulation. Guided by mechanistic insights into CRBN-MGD complex conformational plasticity, we rationally designed indazolone architectures as novel CRBN ligands, developing diverse MGDs with potent ligase binding and exceptional substrate programmability. This platform enables fine-tuned control over neo-substrate recognition, supporting diverse degradation profiles ranging from the broad targeting of critical proteins (IKZF1/3, ZFP91, and LIMD1) to the exquisitely selective degradation of CK1 and IKZF2. Beyond expanding the chemical landscape for intractable targets, our indazolone-based MGDs exhibit favorable pharmacokinetic properties, offering robust promise for therapeutic development. Collectively, the remarkable tunability and exceptional neo-substrate programmability of the indazolone-based platform provide a highly transformative blueprint for next-generation MGD discovery, enabling systematic exploration of the vast CRBN-accessible proteome. Furthermore, as superior CRBN ligands, this platform holds immense potential for PROTAC discovery to systematically address previously intractable targets.

biochemistry↗

bioRxiv2026

Strigolactone signaling regulates corm development through SPL15-mediated hormonal crosstalk in banana

Strigolactones (SLs) are an important class of plant hormones that play crucial roles in regulating plant branching, root architecture, and organ development. However, the regulatory mechanisms underlying the crosstalk between SLs and other plant hormones remain largely unclear, particularly regarding the key regulatory genes that integrate and coordinate multiple hormonal signaling pathways. In this study, secondary cup seedlings of the Pisang Awak banana cultivar Yufen 6 at the eight-leaf stage were used as experimental materials. The roots were treated with a nutrient solution containing 30 mol/L exogenous SLs, while a nutrient solution supplemented with water served as the control. Tissues near the corm growth point were collected at 0, 15, 30, 60, 90, and 120 days after treatment to measure corm weight, height, and diameter, and transcriptome sequencing was performed using the collected tissues. Differentially expressed genes (DEGs) at different treatment stages were identified, followed by Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to systematically investigate the crosstalk between SLs and endogenous hormone metabolism and signaling during corm development in Pisang Awak banana. The results showed that SL treatment significantly inhibited the weight, height, and diameter of the corm. The regulatory effect of SLs on Pisang Awak banana corm development exhibited a clear temporal dynamic pattern, representing a gradual accumulation process that ultimately triggers key developmental transitions. The highest number of DEGs was detected at 15 days after treatment, including 3943 upregulated genes and 3704 downregulated genes, indicating that this stage represents a critical phase for SL response initiation. GO enrichment analysis revealed that the DEGs were mainly involved in metabolic processes, biological regulation, response to stimulus, and regulation of biological processes. KEGG pathway analysis indicated that these DEGs were significantly enriched in pathways related to plant hormone signal transduction, starch and sucrose metabolism, and secondary metabolite biosynthesis. Further analysis revealed that the crosstalk between SLs and multiple hormone metabolic and signaling pathways is mediated by the SPL15 gene, involving auxin (IAA), cytokinin (CTK), abscisic acid (ABA), brassinosteroids (BRs), gibberellins (GA), and jasmonic acid (JA) pathways. This study reveals the molecular mechanism by which SLs regulate Pisang Awak banana corm development through SPL15-mediated integration of multiple hormonal signals, providing new insights into the role of SLs in regulating the development of underground organs in banana.

bioinformatics↗

bioRxiv2026

Telomere-to-telomere gap-free genome assembly integrated with multi-omics uncovers shading mediated regulation of leaf aroma biosynthesis in aromatic crop Pandanus amaryllifolius

Pandanus amaryllifolius is a valuable aromatic crop widely cultivated in tropical forest understories. Here, A telomere-to-telomere, gap-free genome assembly of the Chinese cultivar Taishan Banlan was generated, and the molecular basis underlying shading-mediated regulation of leaf aroma biosynthesis was elucidated for the first time through integrated genomic, transcriptomic, and metabolomic analyses. Volatile metabolomics identified 1599 volatile organic compounds in leaf tissue, with terpenoids as the primary contributors to leaf aroma. Shading modulated the aroma quality of P. amaryllifolius primarily via terpenoid biosynthesis, yielding aroma profiles dominated by sweet, fruity, green, and woody notes. Integrated transcriptomic profiling revealed that terpenoid biosynthesis is co-regulated by structural genes in the MVA/MEP pathways, TPS family members, and key transcription factors, which collectively modulate the biosynthesis and accumulation of the three core differential terpenoids. Our findings provide valuable resources for the targeted genetic improvement and efficient cultivation optimization of P. amaryllifolius.

genomics↗

bioRxiv2026

Reactivating conflicting evaluative memories during sleep reduces decision ambivalence

Memory guides everyday evaluations and decision-making. Yet people often encounter inconsistent information about the same target, giving rise to conflicting evaluative memories and decision ambivalence. Decision ambivalence is not only aversive but also reduces confidence, increases hesitation, and leads to maladaptive choices. While sleep consolidates memories, its role in resolving these evaluative conflicts and shaping decision dynamics remains unknown. Here, we investigated how memory reactivation during sleep, a critical period for memory consolidation and transformation, would reconstruct conflicting evaluative memories and thus influence next-day decision ambivalence. In a valence reversal learning procedure, participants first encoded positive or negative cue-outcome associations (A-B) on Day 1, followed by learning A-C associations yet with opposite valences on Day 2 (i.e., Day 1 negative-to-Day 2 positive and Day 1 positive-to-Day 2 negative). During subsequent non-rapid eye movement (NREM) sleep on Day 2 night, half of the cues were re-presented to sleeping participants to reactivate the cue-associated memories. Upon waking up, participants completed post-sleep evaluation and memory tests. Our results showed that cueing reduced decision ambivalence, especially in the negative-to-positive condition, as evidenced by less curved mouse-tracking trajectories. Meanwhile, cueing promoted the integration of conflicting evaluative memories, again in the negative-to-positive condition. Critically, cueing-induced ambivalence reduction was evident only for items that were integrated after sleep. Electrophysiologically, stronger cue-elicited delta power during NREM sleep predicted next-day ambivalence reduction, while higher cue-elicited spindle probabilities were associated with better memory integration. Together, our findings suggest that memory reactivation during post-learning NREM sleep actively reorganizes conflicting memories, providing a mechanistic pathway through which offline memory reprocessing resolves waking decision ambivalence.

neuroscience↗

bioRxiv2026

Instantaneous Beta Frequency Regulates Self-Generated Timing in Humans

Self-generated timing requires the brain to determine when to terminate an ongoing action without reference to external cues. Although beta-band oscillations have been implicated in temporal control, how their intrinsic dynamics shape moment-to-moment variability in internally generated time remains unclear. Here, combining scalp EEG with intracranial recordings in humans, we show that the instantaneous frequency of beta oscillations (13-30 Hz) provides a dynamic control signal for self-generated timing. During a self-paced time production task, higher instantaneous beta frequency reliably predicted longer produced durations, both within individuals on a trial-by-trial basis and across individuals. This relationship was beta-specific, frontoparietally distributed, robust to motor and spectral confounds, and validated by convergent intracranial recordings. These findings run counter to classical pacemaker-accumulator models and instead establish beta-frequency dynamics as a mechanism that delays action termination by increasing the density of intrinsic stabilizing control updates during self-generated timing.

neuroscience↗

bioRxiv2026

Neurocognitive deficits in controlling aversive memory among insomnia disorders

BackgroundInsomnia disorder is a common sleep disturbance characterized by adverse daytime cognitive and emotional impairments, such as repetitive negative thinking and increased psychological distress. Memory control, a key self-regulatory ability to control or inhibit unwanted thoughts and memories, plays an essential role in supporting cognitive functions and emotional well-being. Here, we delineate the neurocognitive mechanisms underlying memory control among individuals with insomnia. Methods41 participants meeting DSM-5 criteria for insomnia disorder and 40 healthy sleepers completed an emotional Think/No-Think task, during which participants either retrieved (Think) or suppressed the retrieval (No-Think) of aversive memories in response to memory cues while electroencephalograms were recorded. ResultsLinear mixed model analyses with age and depression scores as covariates showed that participants with insomnia exhibited impaired memory control abilities, as evidenced by reduced suppression-induced forgetting in memory recall when compared to healthy sleepers. Electrophysiologically, healthy sleepers showed enhanced right prefrontal theta power in retrieval suppression than in retrieval, indicating elevated needs of inhibitory control during memory control. In sharp contrast, this difference was absent among those with insomnia. Notably, the greater the severity of insomnia symptoms, the smaller the retrieval vs. retrieval suppression theta power differences across participants, linking inefficient top-down control of unwanted memories with low sleep qualities. ConclusionIndividuals with insomnia showed impaired memory control of aversive memories and aberrant electrophysiological activities during retrieval suppression. Future research shall investigate the causal relationship between memory control abilities and insomnia symptoms.

neuroscience↗

bioRxiv2026

STCS: A Platform-Agnostic Framework for Cell-Level Reconstruction in Sequencing-Based Spatial Transcriptomics

Sequencing-based spatial transcriptomics technologies, including Visium HD and Stereo-seq, now enable transcriptome-wide profiling at subcellular resolution. However, these platforms generate measurements over spatially barcoded units rather than biologically segmented cells, creating a fundamental bottleneck for cell-centric analysis and interpretation. Robust recon-struction of coherent single-cell transcriptomes from high-density spatial bins remains an unresolved computational challenge. Here we present STCS (Spatial Transcriptomics Cell Segmentation), a platform-agnostic framework that reconstructs cell-level gene expression profiles by integrating nuclei segmentation with a joint transcriptomic-spatial distance model. STCS is governed by two interpretable parameters and incorporates a reference-free parameter selection strategy based on internal stability and spatial coherence metrics, enabling adaptable deployment across tissue types and technologies without requiring matched ground-truth annotations. We benchmark STCS on a Visium HD human lung cancer dataset with matched Xenium-derived cell segmentation, enabling direct cell-level validation, and on high-resolution Stereo-seq mouse brain data to assess cross-platform generalizability. Across multiple evaluation dimensions--including cell-type agreement, spatial organization, gene-expression fidelity, and compositional accuracy--STCS achieves consistent improvements over existing methods while preserving biologically coherent spatial structure. As sequencing-based spatial transcriptomics is rapidly adopted across biomedical research, STCS provides a broadly applicable and open-source solution for reconstructing cell-resolved transcriptomes, facilitating more reliable downstream analyses and cross-platform integration.

bioinformatics↗

bioRxiv2026

Sex-specific Genetic Regulatory Effects in Chickens

Sexual dimorphism is a defining vertebrate feature, yet its sex-specific molecular architecture remains poorly understood. Here we established a sex-balanced, uniformly reared chicken cohort to map this landscape, integrating individual whole-genome sequencing with 7,969 bulk and 779,380 single nucleus transcriptomes across 32 tissues from 280 birds. We identified 495,098 independent expression quantitative trait loci for 20,194 genes, including 10,937 loci modulated by cell-type composition. Notably, 340 genes were regulated by 449 loci in a sex-dependent manner, significantly enrichment in endocrine tissues like adipose and the adrenal gland. Furthermore, we fine-mapped 1,219 structural variants, demonstrating their unique roles to tissue- and sex-specific expression beyond SNPs. Ultimately, we showed the utility of these regulatory effects in elucidating the molecular basis of metabolism and complex traits in both chickens and humans. This comprehensive atlas of regulatory effects provides profound insights into the genomic and molecular basis of sexual dimorphism in vertebrates.

genomics↗

bioRxiv2026

Pharmacological Inhibition of SLC33A1 Promotes Endoplasmic Reticulum Hyperoxidation and Induces Adaptive IRE1/XBP1s Signaling

The endoplasmic reticulum (ER) transporter solute carrier family 33 member 1 (SLC33A1) has emerged as an attractive therapeutic target in etiologically diverse diseases, ranging from lung cancer to neurodegenerative disorders. Yet, no pharmacologic SLC33A1 modulators have been described. Here, we show that the small molecule IXA4, a highly selective activator of the adaptive IRE1/XBP1s signaling arm of the unfolded protein response (UPR), binds to SLC33A1 and inhibits its activity. Genetic depletion of SLC33A1 phenocopies the selective induction of IRE1/XBP1s signaling brought about by IXA4 treatment. Chemoproteomic analyses and cryo-electron microscopy show that IXA4 binds SLC33A1 within the central channel to inhibit transport of its substrate metabolite(s). Binding of IXA4 to SLC33A1 leads to the accumulation of oxidized glutathione within the ER, hyperoxidizing the ER lumen and inducing activation of adaptive IRE1/XBP1s signaling. Consistent with this function, we find that pharmacologic inhibition of SLC33A1 with IXA4 selectively reduces viability of KEAP1-deficient lung adenocarcinoma cells that have elevated levels of glutathione, mimicking the sensitivity of these cells to genetic deletion of SLC33A1. Our work demonstrates a new physiologic role of SLC33A1 in regulation of ER redox homeostasis and designates IXA4 as a pharmacologic inhibitor of SLC33A1 that can be used to evaluate the biological impact and therapeutic utility of SLC33A1 inhibition in homeostasis and in disease.

cell biology↗

bioRxiv2026

ATLAS: Population-Level Disease Locus Discovery via Differential Attention in Genomic Language Models

Identifying disease-associated genetic variants remains a key challenge in genomics, especially in small cohorts or for rare and complex mutation types where genome-wide association studies (GWAS) often fall short. We introduce ATLAS, a population-level framework that leverages attention signals from pretrained genomic language models (gLMs) to detect disease-associated genes and loci directly from raw sequences--without requiring explicit variant calls or supervised training. ATLAS first performs gene-level differential attention analysis to prioritize candidate genes, followed by base-level analysis to localize disease-associated regions at single-haplotype resolution. We validate ATLAS on synthetic and {beta}-thalassemia datasets, demonstrating robust performance across diverse allele frequencies (down to 10%), cohort sizes (below 200 individuals per group), and genomic scales. Compared to GWAS, ATLAS achieves higher recall of known loci and captures haplotype-specific signals missed by traditional methods. Cross-model benchmarking further shows that precise localization depends on both model size and pretraining on diverse human genomes. In summary, ATLAS offers a scalable, sequence-native alternative to traditional statistical genetics.

genetics↗