Biology subjects

Chu, Y.

Publications and source records attributed to Chu, Y..

At least 19 recordsLinked to original sources

bioRxiv2026

Gastric resection and restriction yield distinct neuroendocrine and metabolic outcomes in obese mice

Whether the metabolic benefits of sleeve gastrectomy arise from gastric restriction alone or also depend on tissue resection remains unclear. We compared sleeve gastrectomy (SG) with a surgical sleeve gastroplasty (GP) model that restricts the stomach without removing tissue. Diet-induced obese male C57BL mice were randomized to SG, GP, or sham surgery and evaluated over six weeks for energy balance, glucose regulation, hepatic lipid accumulation, circulating hormones and branched chain amino acids (BCAAs), and hypothalamic neuropeptide expression. SG and GP produced comparable weight loss of 12 and 11 percent, respectively, without significant differences in energy expenditure. However, the procedures differed in the timing of reduced food intake, and only SG produced smaller, more frequent meals. Both procedures improved glucose handling, with greater improvement after SG. SG also selectively reduced hepatic triglyceride content and circulating BCAAs. Hypothalamic POMC expression decreased after both procedures, whereas NPY and AgRP expression decreased only after SG. SG further produced a distinct hormonal profile, including higher fasting GLP1 and PYY, lower ghrelin, and greater glucose-stimulated increases in several gut and pancreatic hormones. Thus, comparable short-term weight loss after gastric resection and non-resection restriction was accompanied by distinct metabolic and neuroendocrine responses. These findings support a contribution of gastric resection beyond restriction alone and identify candidate gut-brain and metabolic pathways for developing less invasive interventions that reproduce the broader benefits of sleeve gastrectomy.

molecular biology↗

bioRxiv2026

α-Synuclein Seeding in Dopaminergic Neurons Transplanted into the Putamen of Parkinson's Disease Patients

Healthy embryonic neurons grafted into the putamen of Parkinson's disease (PD) patients start to develop Lewy pathology 10 years after transplantation. It remains unclear whether this pathology is initiated by the spread of host-derived -synuclein (syn) aggregates which seed aggregation in the grafted cells or arises independently from a hostile PD environment (e.g., chronic inflammation). To distinguish these possibilities, we identified syn seeding sites in the human brain using a multiplex in situ seed immunodetection (isSID) assay. In PD and dementia with Lewy bodies (DLB), isSID labeled a subset of Lewy pathology (LP) as puncta along neurites and within inclusions. In multiple system atrophy (MSA), isSID labeled Papp-Lantos bodies and microglia. In progressive supranuclear palsy (PSP), isSID labeled tufted astrocytes. isSID labeled neuromelanin in the substantia nigra independent of diagnosis. In 16- and 27-year intraputamenal neuronal grafts, isSID labeled within and proximal to Lewy pathology, as well as outside the grafted tissue. Younger grafts (18 months and 4 years) lacked Lewy pathology. In the 18-month graft, isSID labeled grafted neurons; at 4 years, isSID was confined to amorphous deposits in the graft and was absent from tyrosine hydroxylase-positive grafted neurons. In conclusion, syn seeding is detectable in grafted neurons before and during LP formation. Because seeding accompanies melanization, but is not explained by melanization alone, these findings support host-derived seeds as a contributor to graft pathology.

neuroscience↗

bioRxiv2026

Mutation Order and Selection Shape Intratumor Heterogeneity in Tumor Evolution

Cancer progression often requires the accumulation of multiple driver mutations, yet the same set of drivers may be acquired in different orders. How these alternative mutation-order pathways jointly shape tumor clonal structure remains unclear. We develop a multitype branching-process model in which malignant transformation requires two driver mutations and distinguish malignant cells both by their mutation order and by the independent transformation event that founded their clone. Under a successive exponential approximation, we establish point-process limits for the pathway-specific clone-size processes and derive a closed-form expression for the limiting expected Simpson's index of the combined malignant population. When the two mutation orders produce malignant cells with the same net growth rate, the limiting index separates into effective pathway weights, determined by mutation rates and birth-death dynamics at preceding stages, and within-pathway concentration terms, determined by intermediate-to-malignant growth-rate ratios. This decomposition shows that a driver's effect on heterogeneity depends critically on when it is acquired. A strong driver acquired early expands the intermediate lineage and increases the supply of independent malignant founders, whereas the same driver acquired at the final transition strengthens the growth and age advantage of early-founded malignant clones. Under additive fitness effects, these counteracting mechanisms can produce a non-monotone relationship between selective advantage and clonal concentration. We further show that threshold-like non-additive fitness effects can generate highly concentrated malignant populations, while order-dependent terminal fitness leads the faster-growing pathway to dominate asymptotically. Together, these results reveal how mutation order, mutational accessibility, selection, and epistasis jointly determine lineage-level intratumor heterogeneity.

cancer biology↗

bioRxiv2026

A T2T Benchmark Reveals How Reference Choice Shapes Human Genome Interpretation

The completion of telomere-to-telomere (T2T) human genomes has expanded the accessible landscape of human genetic variation, yet benchmark resources remain limited to conventional high-confidence regions defined by existing reference frameworks. Here, we generated a near-perfect diploid T2T genome (T2T-LIN) from a Chinese individual and established assembly-based truth sets by comparison with T2T-YAO, an ancestry-matched near-perfect T2T reference genome. The benchmark showed a heterozygous/homozygous SNV ratio of ~2, consistent with expectations under Hardy-Weinberg equilibrium, and enabled genome-wide evaluation of reference-dependent biases. We found that reference choice substantially influences genome interpretation: ancestry-matched linear T2T references provided the most faithful representation of individual genomic variation and enabled more accurate genome reconstruction than unmatched linear, diploid and graph-based references. Benchmarking previously inaccessible repetitive and structurally complex regions revealed substantial limitations of current variant callers that were masked by conventional metrics. The T2T-LIN and YAO-LIN benchmarks establish a T2T-era framework for evaluating reference-dependent genome interpretation and variant discovery across nearly the complete human genome.

genomics↗

bioRxiv2026

PerturbLDM: conditional latent diffusion for modelling single-cell perturbation responses

Single-cell perturbation profiling maps intervention-induced phenotypes, yet experiments measure only a fraction of the perturbation-context space. Learning context-dependent perturbation effects could enable response prediction beyond measured conditions. Here we introduce PerturbLDM, a latent-diffusion framework for conditional generation of single-cell transcriptional responses. Following Tahoe-100M pretraining, it predicted 13,942 held-out combinations of observed drugs, doses and cell lines more accurately than existing methods, with higher matched-control effect correlation than an additive marginal baseline in 95.2% of conditions. The Tahoe-100M-pretrained model was further used to rank PANACEA compounds by pathway similarity, placing shared-mechanism pairs among nearest neighbours. In smaller datasets, Per-turbLDM generated a mid-gestational fetal-colon state with 67% lower gene-wise error than Squidiff, retaining the balance between absorptive and BEST4/OTOP2-like epithelial programmes. In PBMCs, it captured six of seven interferon and antiviral programmes and the interferon-associated FAO-OXPHOS programme more accurately than scGen. Together, these results support conditional response generation across data scales and biological settings.

bioinformatics↗

bioRxiv2026

Mtf2 Safeguards Naïve Pluripotency by Restricting Trophectoderm Competence

The first fate decision in mammalian embryos segregates embryonic and extraembryonic lineages, establishing a largely irreversible transition from totipotency to pluripotency. The mechanisms that enforce this unidirectionality remain incompletely understood. Here, we identify the PRC2.1 subunit Mtf2 as a key epigenetic barrier that safeguards lineage fidelity in naive embryonic stem cells (ESCs). Despite globally elevated H3K27me3 levels in naive ESCs, loss of Mtf2 derepresses trophectoderm (TE) regulators, enhances cellular plasticity, generating both ICM- and TE-like populations. Mtf2-deficient cells readily acquire expanded potential with extraembryonic competence and form blastoids without trophoblast stem cell supplementation, whereas loss of the PRC2.2 subunit Jarid2 has minimal effects. Integrated CUT&Tag and transcriptomic analyses reveal that Mtf2 directly represses TE-associated genes in naive ESCs. Thus, reduced Mtf2 occupancy permits their activation in expanded potential states. Collectively, our findings establish Mtf2 as a central epigenetic barrier that stabilizes naive pluripotency and restricts inappropriate acquisition of extraembryonic fate.

Developmental Biology↗

bioRxiv2026

Conditioned pain modulation recruits the descending pain modulatory system to inhibit spinal activity

Conditioned pain modulation (CPM) is a phenomenon where the perception of pain is reduced in the presence of a second painful stimulus. In rodents, a similar phenomenon involves descending control pathways that rely on the brainstem to modulate spinal cord activity while in humans, descending mechanisms including the spinal cord remain elusive. In this study, we applied functional MRI to simultaneously record brain and spinal cord responses during CPM in healthy participants. CPM developed over time and was associated with reduced activity in key pain-related brain regions and the spinal cord dorsal horn. Conversely, regions involved in descending pain modulation showed increased activity during CPM. Functional connectivity analyses revealed reduced coupling during CPM between spinal and descending control regions. Our findings demonstrate that CPM involves both the descending pain control system to inhibit spinal activity and bottom-up pain signalling of attenuated spinal responses.

neuroscience↗

bioRxiv2025

Swine wastewater drives dissemination of high-risk blaNDM-positive Escherichia coli clones

Swine wastewater can accelerate the spread of antibiotic-resistant bacteria among animals, the environment, and humans. Therefore, the molecular epidemiological characteristics of carbapenemase-producing E. coli isolates from swine wastewater was investigated. In this study, a total of 100 no-cloning carbapenemase-producing (blaNDM-1, blaNDM-5 and blaOXA-48-like) E. coli isolates were recovered from 316 swine wastewater samples collected from 29 swine farm in Shandong, China. Of which, two isolates that co-harbored blaOXA-48-like and mcr-1, as well as five isolates that co-harbored blaNDM-5 and tet(X4). All isolates were multidrug-resistant (MDR), with a majority exhibiting resistance to meropenem, ciprofloxacin, imipenem and gentamicin. Whole genome sequencing (WGS) analysis indicated that these isolates were belonged to 12 distinct sequence types (STs), with the most prevalent STs being ST10 and ST5299. Additionally, Phylogenomic analysis revealed that clonal spread of blaNDM-positive ST10 E. coli isolates at swine wastewater in multiple cities, including Binzhou, Dezhou, Heze and Weihai. Meanwhile, a notable divergence in SNPs was observed in isolates from this study and public database, which indicates that these blaNDM-positive E. coli isolates have high genetic diversity. In addition, WGS analysis further revealed that blaNDM co-existed with other antibiotic resistance genes, conferring resistance to twelve classes of antimicrobials. This study underscores the importance of surveillance for blaNDM-harboring microbes in swine wastewater, and continuous monitoring in swine wastewater is essential for ensure human and environmental health. IMPORTANCEThe study described the NDM-positive E. coli strains isolated from swine wastewater, where it was associated with antibiotic exposure. Notably, swine wastewater was increasingly considered to be related to CRE hosts. Pathogens originating from this source can be transmitted to humans via direct and indirect contact, as well as through the ingestion or inhalation of contaminated water or aerosols. In addition, under the selective pressure of antibiotics, the co-harbored of blaOXA-48-like and mcr-1, blaNDM-5 and tet(X4) genes within the same host bacterium signifies the emergence of an extremely MDR pathogen, posing a grave threat to public health. Consequently, the monitoring of blaNDM in swine wastewater is critical for containing the transmission and persistence of carbapenem-resistant pathogens.

microbiology↗

bioRxiv2025

Longitudinal analysis of bacterial community dynamics in normal hospital ward following patient occupancy

The hospital microbiome significantly influences patient recovery and clinical outcomes. However, the dynamics of microbial colonization and transmission following initial patient occupancy remain poorly understood. Here, we employed 16S rRNA gene amplicon sequencing to investigate bacterial community dynamics on surfaces within neurosurgery ward and patients as a new hospital became operational. Our results showed that bacterial colonization in hospital wards follows distinct site-specific patterns, after hospital opening, alpha diversity was significantly increased on floors and drawer handles but decreased on bedrails and faucet handles compared to preopening. Beta diversity analysis showed that surfaces frequently contacted by patients exhibited the greatest compositional turnover, such as bedrails, drawer handles, and faucet handles, bacterial communities in after-opening were more homogeneous across sites than preopening, indicating potential bacterial transmission. Moreover, we found following patient admission, patient hand-derived microbiomes exert a significant influence on the bacterial communities in hospital wards, with a particularly pronounced impact on bedrails. Additionally, the potential disease risk of bedrails in post-opening was significantly higher than preopening. Taken together, these findings underscore the critical role of human contact in shaping hospital microbiomes and highlight the importance of targeted infection control strategies to mitigate pathogen transmission.

bioinformatics↗

bioRxiv2025

A complete diploid human genome benchmark for personalized genomics

Human genome resequencing typically involves mapping reads to a reference genome to call variants; however, this approach suffers from both technical and reference biases, leaving many duplicated and structurally polymorphic regions of the genome unmapped. Consequently, existing variant benchmarks, generated by the same methods, fail to assess these complex regions. To address this limitation, we present a telomere-to-telomere genome benchmark that achieves near-perfect accuracy (i.e. no detectable errors) across 99.4% of the complete, diploid HG002 genome. This benchmark adds 701.4 Mb of autosomal sequence and both sex chromosomes (216.8 Mb), totaling 15.3% of the genome that was absent from prior benchmarks. We also provide a diploid annotation of genes, transposable elements, segmental duplications, and satellite repeats, including 39,144 protein-coding genes across both haplotypes. To facilitate application of the benchmark, we developed tools for measuring the accuracy of sequencing reads, phased variant call sets, and genome assemblies against a diploid reference. Genome-wide analyses show that state-of-the-art de novo assembly methods resolve 2-7% more sequence and outperform variant calling accuracy by an order of magnitude, yielding just one error per 100 kb across 99.9% of the benchmark regions. Adoption of genome-based benchmarking is expected to accelerate the development of cost-effective methods for complete genome sequencing, expanding the reach of genomic medicine to the entire genome and enabling a new era of personalized genomics.

genomics↗

bioRxiv2025

Gene-Family Encoding Boosts Domain-Adapted Single-Cell Language Models

Transformer-based single-cell foundation models often rely on ranked-gene (RG) sequences where genes, ranked by expression, are often not functionally related, weakening next-token learning and the structure of learned embeddings. Here, we introduce gene-family (GF) encoding, where expressed genes are grouped into functionally defined families, and ranking performed within each family. Using 100,000 gastric-cancer (GC) cells, we domain-adapted 8-billion-parameter Llama and Qwen backbones with either RG or GF sentences and benchmarked zero-shot performance on embedding-based and generative tasks. GF models outperformed RG models across both tasks and backbones. We scaled GF-Llama to 1.3 million GC cells to obtain GF-Llama-GC and applied it to two applications: resolving fine-grained cellular heterogeneity and discovering cell populations associated with disease progression. GF-Llama-GC revealed immune-cell subclusters not resolved by standard expression-based analyses. Applying in-silico cell removal/transplantation on a chemotherapy responder/non-responder scRNA-seq dataset, GF-Llama-GC highlighted not only epithelial cells but also neutrophils as key cells associated with chemotherapy response.

bioinformatics↗

bioRxiv2025

Approaching an Error-Free Diploid Human Genome

Complete human genomes now resolve regions long inaccessible to genetic analysis, yet repetitive and structurally complex sequences remain prone to assembly errors because of limitations in current assembly algorithms. We introduce a support-based framework founded on the principle that each sequencing read provides independent evidence for the underlying genome sequence. Implemented through Sufficient Alignment Support (SAS), the framework identifies regions lacking concordant read support, localizes residual errors, and guides correction. Applying SAS to T2T-YAO, a haplotype-resolved Han Chinese genome, produced a support-validated assembly in which nearly all sequences outside unresolved rDNA arrays and long homopolymer tracts are supported by independent sequencing evidence. This work establishes a scalable framework for genome validation and provides a support-validated East Asian diploid reference for investigating human genomic diversity.

genomics↗

bioRxiv2025

Impact of Inactivation Methods on Biosafety and Antigen Reactivity of Brucella melitensis from the Perspective of Astral-DIA Proteomics Based on Antibody Immunoprecipitation Mass Spectrometry

Effective Brucella inactivation is imperative for safe vaccine development, diagnostics, and sample handling, particularly in resource-limited regions lacking high-level containment facilities. This study investigated inactivation methods using the Rev.1 vaccine strain and three Brucella melitensis field isolates from Gansu (GS-XG, GS-SN, GS-MQ). Heat (80{degrees}C/95{degrees}C, 10-20 min) and formaldehyde (0.4%/0.6%, 48-72 h) inactivation were evaluated for biosafety and antigenicity. Rev.1 was completely inactivated by 80{degrees}C/95{degrees}C for 20 min or 0.6% formaldehyde for 48-72 h. Superior antigenicity, compared to phenol inactivation, was confirmed by ELISA and Western Blot. However, field isolates demonstrated greater resistance, surviving 80{degrees}C for 20 min and 0.4% formaldehyde for 72 h, necessitating stricter conditions (95{degrees}C, 20 min; 0.6% formaldehyde, 72 h) for their complete inactivation. Astral-DIA proteomics analyzed approximately 60% of the proteome ([~]2000/3300 proteins), revealing 256, 311, and 318 differentially expressed proteins between 80{degrees}C/95{degrees}C heat, 48-h/72-h formaldehyde, and heat/formaldehyde methods, respectively. Gene Ontology and KEGG analyses indicated heat inactivation upregulated cellular structure proteins but downregulated metabolic pathways, with 95{degrees}C potentially damaging conformational epitopes. Formaldehyde treatment (48 h) stabilized soluble antigens, preserving ribosomal and regulatory protein epitopes, while 72-h treatment induced organelle disintegration. Protein-Protein Interaction networks suggested heat inactivation enhanced immunogenicity of membrane and stress proteins, suitable for targeted studies, whereas formaldehyde preserved broader epitopes, beneficial for whole-cell vaccines and multi-epitope screening. Inactivation methods must be tailored to specific strain characteristics and applications. Astral-DIA provides molecular insights into antigenicity loss, guiding future research on protein functions and epitope dynamics for precise brucellosis control.

microbiology↗

bioRxiv2025

Semantics across the globe: A universal neurocognitive semantic structure adaptive to climate

Thousands of languages are used worldwide as the primary means of human thought communications. While both similarities and variations in word meaning (semantics) across different languages are well recognized, the underlying mechanisms remain enigmatic without a coherent theoretical model for semantic representation. Given that semantic representation is a product of the human brain, we address this issue through the lens of neurocognitive theories, with the consensus framework that semantics are derived from sensory experiences, with a set of dimensions being identified as biologically salient in neuroscientific studies. We operationalized word semantic representations with this set of specific dimensions, using computational models (53 languages word embedding data; Study 1), human behavioral ratings (253 subjects, 8 languages; Study 2), and brain activity data (86 subjects, 45 languages; Study 3), and analyzed the similarity and variation patterns of concepts across different languages. These three approaches converge on the finding that, across diverse language samples, word semantic representations along the neurocognitive dimensional structures exhibit strong commonalities, with variations along this structure being significantly and uniquely explained by climate, beyond sociocultural-centered variables. These results present a universal, biologically constrained semantic structure that is adaptive to environmental inputs, reconciling the classical universality and relativity debate.

neuroscience↗

bioRxiv2025

Age-Dependent Patterns of Gut Bifidobacterium Microbiome Diversity across child Cohorts in a Multi-Ethnic population, Western China

High-throughput sequencing has provided unprecedented opportunities to elucidate the role of human gut microbiota. However, current studies have primarily focused on infant and adult cohorts, leaving biomarkers associated with the development of transitional children unclear, particularly in revealing geographic, multi-ethnic group underrepresented. Here, we analyzed fecal samples from 776 infants and children aged 0 to 15 years with amplicon sequencing of the 16S rRNA and groEL gene, representing multi-ethnic groups from different residential locations in the Yining region of Xinjiang. First, we systematically characterized the gut microbiota Spectrum of children in the Ili region. As age increases, the diversity and functional composition of the microbiota become richer and more stable, with Bifidobacterium consistently dominating the community. Utilizing Bayesian regression models, we identified a cluster of microbiota strongly associated with growth and development in children, which phylogenetically corresponds to Bifidobacterium breve. Finally, we included another similar study in the species-level of Bifidobacterium, and metrics of the age-related effect sizes showed heterogeneity across ethnic groups (Uyghur vs Han: {beta}Deming = 4.40 and Kazakh vs Han: {beta}Deming = 4.33). In conclusion, this study provides novel insights into the developmental trajectories of the gut microbiome in children and its relationship with health. IMPORTANCEThe intestinal microbial markers associated with the development of geographically diverse, multi-ethnic children, particularly those represented by Bifidobacterium species, have been elucidated in this study. Through the application of Bayesian regression models, we identified a microbiome profile closely linked to childhood growth and development, specifically phylogenetically aligned with Bifidobacterium breve. When incorporating a comparable study conducted at the species level of another Bifidobacterium, it was observed that age-related effect sizes varied among different ethnic groups, offering novel insights into the developmental trajectory of childrens gut microbiota and its implications for health.

microbiology↗

bioRxiv2025

Axon Trafficking Counteracts Aberrant Protein Aggregation in Neurons

Directed axon trafficking of mRNA via ribonucleoprotein complexes (RNPs) is essential for the proper function and survival of neurons. However, the mechanisms governing RNP transport in axons remain poorly understood. Here, we identify Annexin A7 (ANXA7) as a critical adaptor facilitating the retrograde transport of T-cell intracellular antigen 1 (TIA1)-containing RNPs by linking them to the cytoplasmic dynein. Persistent axonal Ca{superscript 2} elevation disrupts ANXA7s linker role, causing the detachment of TIA1 granules from dynein, consequently impairing transport and triggering pathological TIA1 aggregation within axons. Similarly, ANXA7 knockdown decouples TIA1 granules from dynein, severely obstructing trafficking and causing pathological aggregation of TIA1 in axons, which culminates in axonopathy and neurodegeneration both in vitro and in vivo. Conversely, ANXA7 overexpression enhances trafficking and counteracts aberrant aggregation of TIA1-containing RNPs in axons. Our findings elucidate a novel mechanism underlying RNP axonal transport, highlighting its significance in the biology and pathology of central neurons.

neuroscience↗

bioRxiv2024

WRAD core perturbation impairs DNA replication fidelity promoting immunoediting in pancreatic cancer

It is unclear how cells counteract the potentially harmful effects of uncoordinated DNA replication in the context of oncogenic stress. Here, we identify the WRAD (WDR5/RBBP5/ASH2L/DPY30) core as a modulator of DNA replication in pancreatic ductal adenocarcinoma (PDAC) models. Molecular analyses demonstrated that the WRAD core interacts with the replisome complex, with disruption of DPY30 resulting in DNA re-replication, DNA damage, and chromosomal instability (CIN) without affecting cancer cell proliferation. Consequently, in immunocompetent models, DPY30 loss induced T cell infiltration and immune-mediated clearance of highly proliferating cancer cells with complex karyotypes, thus improving anti-tumor efficacy upon anti-PD-1 treatment. In PDAC patients, DPY30 expression was associated with high tumor grade, worse prognosis, and limited response to immune checkpoint blockade. Together, our findings indicate that the WRAD core sustains genome stability and suggest that low intratumor DPY30 levels may identify PDAC patients who will benefit from immune checkpoint inhibitors.

cancer biology↗

bioRxiv2024

scEMB: Learning context representation of genes based on large-scale single-cell transcriptomics

BackgroundThe rapid advancement of single-cell transcriptomic technologies has led to the curation of millions of cellular profiles, providing unprecedented insights into cellular heterogeneity across various tissues and developmental stages. This growing wealth of data presents an opportunity to uncover complex gene-gene relationships, yet also poses significant computational challenges. ResultsWe present scEMB, a transformer-based deep learning model developed to capture context-aware gene embeddings from large-scale single-cell transcriptomics data. Trained on over 30 million single-cell transcriptomes, scEMB utilizes an innovative binning strategy that integrates data across multiple platforms, effectively preserving both gene expression hierarchies and cell-type specificity. In downstream tasks such as batch integration, clustering, and cell type annotation, scEMB demonstrates superior performance compared to existing models like scGPT and Geneformer. Notably, scEMB excels in silico correlation analysis, accurately predicting gene perturbation effects in CRISPR-edited datasets and microglia state transition, identifying a few known Alzheimers disease (AD) risks genes in top gene list. Additionally, scEMB offers robust fine-tuning capabilities for domain-specific applications, making it a versatile tool for tackling diverse biological problems such as therapeutic target discovery and disease modeling. ConclusionsscEMB represents a powerful tool for extracting biologically meaningful insights from complex gene expression data. Its ability to model in silico perturbation effects and conduct correlation analyses in the embedding space highlights its potential to accelerate discoveries in precision medicine and therapeutic development.

bioinformatics↗