Biology subjects

Wan, S.

Publications and source records attributed to Wan, S..

At least 19 recordsLinked to original sources

bioRxiv2026

Myelinated mechano-nociceptors encode noxious cold and underlie hypersensitivity in pulpitis

The ability to identify dangerous temperatures is critical, but the neural basis for cold sensation remains enigmatic. By combining in vivo functional imaging with tissue-specific spatiotemporal cooling, we unveil myelinated high threshold-mechanonociceptors (HTMRs) as the neural origin of tooth cold sensitivity. Critically, this neural encoding is driven by threshold noxious cold temperature, not cooling rate and remains independent of the cool receptor TRPM8 or pain transducer Nav1.8. Furthermore, while Bmpr1b-Cre captures the intradental HTMRs, it further reveals myelinated HTMRs mediate trigeminal cold encoding in soft tissues. Finally, we determine that inflammation potentiates intradental HTMRs, providing functional basis for clinical hypersensitivity in pulpitis. Collectively, our results reveal a new neural mediator of cold sensation and provide insight into painful transformation of cold sensory encoding in inflammation.

neuroscience↗

bioRxiv2026

DAG-HEART: Directed Acyclic Graph-Guided Health Equity-Aware Representation Transfer Learning Framework for Breast Cancer

Breast cancer outcome prediction remains challenging for underrepresented populations because genomic datasets are demographically imbalanced and conventional multi-omics integration largely relies on undirected molecular similarity. We developed DAG-HEART, a directed acyclic graph-guided multi-omics transfer-learning framework that extends our previous transfer learning strategy with data augmentation. Using TCGA-BRCA mRNA, miRNA, and DNA-methylation data, DAG-HEART was evaluated for progression-free interval prediction in a data-minority group. DAG-guided nonlinear integration consistently improved predictive performance relative to direction-agnostic and correlation-based representations, while biologically motivated directional constraints generally outperformed reversed or unconstrained structures. Recurrently selected features converged on extracellular-matrix and regulatory pathways and supported clinically meaningful risk stratification. DAG-HEART provides an interpretable strategy for combining directed multi-omics structure with transfer learning under data imbalance across racial groups.

bioinformatics↗

bioRxiv2026

High-Resolution Subtyping of Pediatric Low-Grade Glioma Using an Integrated Meta-Clustering Framework

Pediatric low-grade glioma (pLGG) is the most common type of brain tumor in children, accounting for approximately 30% of all central nervous system tumors in children. pLGG has multiple molecular subtypes that differ in disease progression, recurrence patterns, and treatment responses. Conventional wet lab approaches including molecular profiling and histopathological studies for pLGG characterization are time consuming, costly, and laborious. Recently, methods based on artificial intelligence (AI) or machine learning (ML) have been widely used for pLGG molecular categorization, but most of them can only identify two or three pLGG subtypes. To more comprehensively characterize the molecular subtypes of pLGG and their potential biological and therapeutic significance, we develop an integrated meta-clustering approach, namely Meta-pLGG, that can explore high resolution molecular subtypes and their transcriptional heterogeneity for pLGG. Specifically, we first performed multiple rounds of random projection (RP) to generate dimension-reduced feature vectors from pLGG transcriptomics data, each of which was subsequently clustered by different clustering algorithms including hierarchical clustering, K-means, Self-Organizing Maps (SOM), Non-negative Matrix Factorization (NMF), Gaussian Mixture Model (GMM), and Spectral Clustering, as base clustering methods. Then, to yield robust clustering performance, we integrated the clustering results of these RP based individual clustering algorithms by adopting a weighted meta-clustering (wMetaC) approach. Results based on 532 pLGG patients suggested that our proposed approach demonstrated superior stability and discriminative powers for higher resolution pLGG subtyping compared to conventional approaches. Based on consensus matrix analysis, we identified two major pLGG mega-subtypes, with one further subdivided into three subgroups and the other into two. Then, we performed cluster specific differential gene expression analysis, molecular pathway analysis, and gene-drug-disease association analysis. The results showed that the identified five subgroups exhibited significant subtype-specific transcriptomic heterogeneity. In summary, our meta-clustering approach demonstrated much higher performance and robustness in identifying higher resolution molecular subtypes of pLGG, revealing the molecular heterogeneity within pLGG and potentially providing new insights for more precise molecular subtyping and precision therapy.

bioinformatics↗

bioRxiv2026

LNP-CpG: deploy the self-adjuvant role of mRNA vaccines

With the rapid advancement of mRNA vaccines, lipid nanoparticles (LNPs) have emerged as pivotal carriers and adjuvants for non-mRNA vaccine modalities, driven by their superior nucleic acid delivery efficiency and intrinsic self-adjuvanting properties. In this study, we systematically evaluated various formulation strategies combining LNPs and the CpG adjuvant within a varicella-zoster virus glycoprotein E (VZV-gE) subunit vaccine framework. We demonstrated that uniform nanoparticles formed by LNP-encapsulated CpG (LNP-CpG), when simply admixed with the gE antigen, elicited superior immunogenicity compared to alternative encapsulation configurations. Intramuscular administration of a two-dose (LNP-CpG)+gE regimen significantly augmented both humoral and cellular immune responses in mice, markedly outperforming the commercial vaccine Shingrix (administered at a 1/10 human dose). Crucially, the identical regimen induced robust, comparable immune profiles to a full human dose of Shingrix in rhesus macaques. Furthermore, LNP-CpG displayed broad-spectrum utility across diverse vaccine platforms, demonstrating efficacy against both respiratory (RSV) and neurotropic (HSV) pathogens, compatibility with multiple modalities, including subunit (VZV-gE, RSV-Pre-F), live-attenuated (LA-HSV), and inactivated (i-HSV) vaccines; and versatile implementation in a combined VZV+RSV formulation. Collectively, our findings position LNP-CpG as a versatile, safe, highly efficacious adjuvant platform with substantial clinical translational potential, offering a compelling paradigm for next-generation vaccine development.

immunology↗

bioRxiv2026

SAMP V2: A novel stacking ensemble learning model for antimicrobial peptides identification based on augmented split amino acid composition with biochemical-sequence-order information

Antimicrobial resistance reduces the effectiveness of conventional antibiotics and has become a major global health threat, highlighting the need for new anti-infective agents. Antimicrobial peptides (AMPs), a diverse class of innate immune effectors with broad-spectrum antimicrobial activity, are promising candidates for combating drug-resistant infections. Identifying AMPs by wet-lab experiments, however, remains costly and time-consuming, creating a strong demand for computational identification methods. Our recently developed method, SAMP, captures region-specific residue distributions based on proportionalized split amino acid composition. However, SAMP might ignore key biochemical information and sequence order information. Here we present SAMP V2, a stacking ensemble learning framework based on biochemical and sequence-order information augmented split amino acid composition (BIA-SAAC), which extends SAMP by integrating pseudo-amino acid composition features with biochemical and sequence-order information into split peptide regions. Specifically, each peptide is divided into N-terminal, middle, and C-terminal regions, and pseudo amino acid composition is calculated within each region. Benchmarking tests on six independent test datasets, SAMP V2 outperformed multiple state-of-the-art models, including AMPpred-MFA and iAMP-Attenpred, in terms of accuracy, MCC, G-measure and F1-score. Given its high and robust performance, SAMP V2 could significantly accelerate the discovery of next-generation antimicrobial therapeutics for addressing the global threat of multidrug-resistant pathogens.

bioinformatics↗

bioRxiv2026

MetaAMI: A Novel Meta-Learning Approach for Predicting In-Hospital Mortality in Acute Myocardial Infarction

Acute myocardial infarction (AMI) is one of the leading cardiovascular diseases worldwide and remains a major cause of mortality. Early risk prediction can help clinicians identify high risk patients shortly after admission and support timely monitoring and individualized treatment. Previous AMI risk assessment approaches predominantly rely on a single model structure or fixed feature representation, which may limit their ability to capture diverse risk related patterns and reduce predictive performance. To address these challenges, we propose MetaAMI, a random projection based meta-learning framework for AMI outcome prediction. Specifically, patients were selected based on ICD-9 and ICD-10 diagnostic codes for AMI from the Medical Information Mart for Intensive Care IV (MIMIC-IV) v3.1 database. Features were transformed through multiple random projections, with each random projection generating a distinct lower dimensional feature representation. Subsequently, baseline classifiers were trained on each lower dimension representation to predict in-hospital mortality among AMI patients. The predictions were aggregated to construct an integrated feature representation, which was used as input to a meta-learner architecture. By effectively integrating complementary information from diverse baseline models, the meta-learner refined the decision boundary and enhanced overall predictive performance. Survival analysis and SHapley Additive exPlanations (SHAP) analysis were further performed to evaluate clinical utility and interpret the model predictions. Benchmarking results based on MIMIC-IV dataset suggested that our MetaAMI consistently outperformed all the baseline classifiers across seven evaluation metrics including Accuracy, Area Under the Curve (AUC), F1 Score, G-Measure, Jaccard Index, Youden J, and Matthews Correlation Coefficient (MCC). In addition, feature importance analysis showed clinically relevant predictors of in-hospital mortality. In summary, MetaAMI provides an effective and robust solution for machine learning based AMI risk prediction. We anticipate that the application of MetaAMI will have a positive impact on clinical risk stratification and personalized treatment strategies for AMI.

bioinformatics↗

bioRxiv2026

Immunometabolic Reprogramming by Thyroid-Stimulating Antibodies Drives Orbital Adipogenesis via Histone Lactylation in Thyroid Eye Disease

Thyroid eye disease (TED) is conventionally viewed as an autoimmune inflammatory disorder. However, the metabolic determinants driving orbital tissue remodeling remain largely undefined. Here, we uncover a paradigm-shifting mechanism whereby thyroid-stimulating antibodies (TSAbs) instigate profound metabolic reprogramming in orbital fibroblasts (OFs), driving a switch toward glycolysis and markedly elevating lactate production via the CREB/LDHA axis. Critically, we demonstrate that this TSAbs-induced metabolic perturbation is not a mere byproduct but the principal driver of pathology. Mechanistically, lactate functions as a signaling metabolite that directly potentiates adipogenesis via histone lactylation-dependent transcriptional activation of the master adipogenic regulators PPAR{gamma} and C/EBP. Collectively, our findings underscore metabolic reprogramming as a central pathogenic mechanism in TED, whereby autoimmune cues drive disease progression through metabolic-epigenetic crosstalk. This work substantiates the paradigm of TED as an immune- metabolic disorder.

immunology↗

bioRxiv2026

Multidimensional variation and population stratification across 8000 complete human centromeres

Human centromeres are indispensable for the faithful segregation of chromosomes during cell division, yet their highly repetitive nature has historically precluded comprehensive characterization, leaving fundamental questions about their sequence diversity, evolution trajectories and function dynamics unresolved. Here, we generated 6,312 complete human centromere sequences from 320 phased genome assemblies in Asian Pan-Genome project phase 1. By integrating the assemblies from the Human Pangenome Reference Consortium (HPRC) and Human Genome Structural Variation Consortium (HGSVC), we constructed a multidimensional genetic variation map encompassing over 8,000 gapless centromeres. Centromeric satellite arrays account for 4.19% to 6.01% of the whole genome, with substantial variations in size and architecture across chromosomes. Using a refined alpha satellite clustering approach that captures global diversity, we identified 195 higher-order repeat (HOR) arrays, 56.4% of which are absent from the T2T-CHM13 reference genome. Extensive structural variations across multiple dimensions exhibit population stratification, including centromeric haplotypes (CenHaps), ultra-large pericentric inversions spanning up to 36.6 Mbp, and inter-chromosomal HOR sharing that reflects sequence exchange among chromosomes. Integrating CENP-A CUT&Tag experiments and long-read-based DNA methylation profiles, we demonstrate that 16.8% of centromeres harbor multiple potential kinetochore assembly sites, and CenHap-specific local HOR homogenization is associated with kinetochore positioning. Despite global suppression of recombination at centromeres, we observed asymmetric linkage disequilibrium flanking centromeres and an ancient recombination event within the centromere of chromosome 19. Furthermore, contrary to the prevailing assumption of high mutation rates in centromeres, our estimates based on stringent orthology reveal no significantly higher single-base substitution rates for centromeres relative to flanking pericentromeric regions with substantial variations across chromosomes, despite extraordinary structural plasticity. Collectively, these multi-scale centromeric variations provide a global view of human centromere diversity and population stratification, fundamentally redefine centromere evolution through a dual-track model balancing structural innovation with mutational constraint, and establish an essential resource for investigating centromere biology and a baseline reference for diagnosing centromere-associated disorders.

genomics↗

bioRxiv2026

Unearthing a fungal giant: Dianjunaceae fam. nov., a novel Paleocene lineage of Xylariales harbouring Dianjunus rex gen. et sp. nov.

During mycological surveys in Yunnan Province, China, specimens of a fungus producing massive, upright stromata up to 50 cm high and individually 2.2 Kg in weight were sampled. Through an integrative taxonomic approach combining detailed morphology, multilocus phylogeny (ITS, LSU, RPB2, TUB2), and phylogenomic analyses, this fungus is proposed as the new species Dianjunus rex gen. et sp. nov., the type of the new family Dianjunaceae (Xylariales). Phylogenetic analyses robustly place Dianjunaceae as a distinct sister clade to Graphostromataceae. Divergence time estimation dates the origin of this family to the early Paleocene ([~]65 Mya), coinciding with the post-K-Pg extinction period, when an estimated 75% of all plant and animal species went extinct, and a significant ecological reorganization of life on earth happened. The stromata of D. rex represent the largest fructifications documented within the Ascomycota, significantly expanding the known morphological range of the Xylariales. The study provides a comprehensive description, including a nodulisporium-like anamorph with periconiella-like branching patterns, and discusses the taxons phylogenetic placement, and distinctive morphology. This discovery highlights the unexplored fungal diversity in East Asian forests.

microbiology↗

bioRxiv2026

Accurate de novo design of peptides from programming biophysical landscape

Peptides regulate virtually every cellular process and emerge as transformative modalities in bioengineering and therapeutics. However, the de novo design of functional peptides remains challenging, as peptide-protein interactions are governed by delicate and dynamic biophysics that are difficult to model, and experimental datasets are scarce. Here we present NeoPep, a generative deep-learning framework that encodes biophysical principles to accurately design functional peptides de novo. By integrating over 5 million peptide-protein complexes spanning experimentally determined, sequence mimics, and structure ensembles, NeoPep learns the complex biophysical landscape governing peptide function and enables its programmable control. In prospective application across 10 diverse and challenging targets, NeoPep generated potent peptide binders, agonists, and antagonists with hit rates of 12.5-66.7%, even in the absence of defined binding sites or structure information. Beyond de novo co-design, NeoPep supports standalone structure or sequence redesign, readily discriminating subtle context differences. In the structure redesign mode, it generates highly selective peptides with atomic-level conformational accuracy (C RMSD < 2.0 [A] to our solved cryo-EM structure). This structural precision circumvents the need for experimental structure determination, further accelerating iterative sequence redesign to yield a 43.3-fold improvement in potency. These findings establish a general framework for translating biophysical principles into actionable peptide functions, with broad implications for basic research, bioengineering, and medicine.

molecular biology↗

bioRxiv2026

Molecular Characterization of T-Lineage Acute Lymphoblastic Leukemia by an Optimal-Transport Based Multi-Omics Integration Framework

T-lineage acute lymphoblastic leukemia (T-ALL) is an aggressive pediatric malignancy characterized by complex heterogeneity across multiple molecular layers. Accurate subtyping is essential for understanding disease mechanisms, risk stratification, and guiding targeted therapeutic strategies. However, current diagnostic approaches are labor-intensive and time-consuming, and existing computational methods are limited by reliance on single-modality data or simple integration strategies. Effective integration of heterogeneous multi-omics data remains a major computational challenge. We present OTTER (Optimal Transport-based Transcriptomics and gEnomics Representation fusion), a novel multi-modal deep learning framework that jointly models RNA-seq gene expression and somatic genomic variant data for T-ALL molecular characterization. OTTER encodes each omics modality through a modality-specific variational autoencoder and aligns the resulting latent representations using Gromov-Wasserstein optimal transport (GW-OT), which preserves the internal geometric structure of each modality without requiring a shared feature space. We applied OTTER to the Childrens Oncology Group (COG) AALL0434 cohort comprising 1,309 patients across 17 T-ALL subtypes. Gradient-based feature importance and cross-omics interaction analysis were performed on the holdout set to identify subtype-driving molecular features and cross-modal coordinated programs. OTTER provides a principled, biologically interpretable, and computationally effective framework for multi-omics-driven T-ALL molecular characterization. By leveraging GW-OT for geometry-preserving cross-modal alignment and gradient-based interpretability for cross-omics interaction profiling, OTTER goes beyond single-modality approaches to uncover the coordinated molecular landscape of T-ALL. The framework is generalizable to other cancers and multi-omics integration tasks.

bioinformatics↗

bioRxiv2026

Atlas-Level Single-Cell and Spatial Transcriptomics Data Integration via PRIME

Single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics (ST) have enabled atlas-scale cellular cartography, with consortium efforts now assembling millions of cells across diverse tissues, donors, and technologies to build comprehensive references for cell identify and disease mechanism, yet the scientific value of these atlases hinges on robust computational integration across heterogeneous data sources. Unlike pairwise batch correction, atlas-level integration must jointly reconcile heterogeneous and often hierarchically nested batch effects across many datasets whose cell-type compositions are highly imbalanced, all while preserving subtle biological variation and remaining computationally tractable at the scale of millions of cells. Existing approaches often prioritize either batch mixing or preservation of local biological structure, and most cannot natively accommodate spatial coordinates. Here we introduce PRIME (Projection-based Robust Integration via Manifold Embedding), an ensemble integration framework that combines random-projection-based consensus anchoring, graph-Laplacian correction, and optional spatial-neighborhood regularization. Across multiple random projections of the expression manifold, PRIME uses consensus voting to keep only cell pairs that repeatedly matched, reducing false anchors caused by projection-specific distortions. For ST, PRIME couples this expression-based anchor graph with a coordinate-derived spatial neighborhood graph in a unified graph-Laplacian objective with closed-form solution, enabling simultaneous cross-batch alignment and local spatial coherence. Based on extensive benchmarking spanning diverse datasets, we show that PRIME consistently outperforms state-of-the-art methods in both batch correction and biological conservation across scRNA-seq and ST integration scenarios and downstream tasks including trajectory inference, spatial-domain preservation, and perturbation-response analysis. Particularly, when integrating a human hematopoiesis benchmark spanning eight donors and approximately 33,000 cells, PRIME preserves biologically coherent developmental trajectories in human hematopoiesis. It also maintains cortical laminar architecture across dorsolateral prefrontal cortex sections in a ST dataset and recovers known drug-target relationships in a perturbation atlas of more than 1 million cells while suppressing batch-associated confounders. Together, these results establish PRIME as a versatile and scalable framework for atlas-level integration of scRNA-seq and ST across diverse biological applications.

bioinformatics↗

bioRxiv2026

Complete Telomere-to-Telomere Assembly of the Y Chromosome in the Chinese Quartet

The complete assembly of the human Y chromosome remains a formidable task due to its highly repetitive, palindromic, and heterochromatic architecture. While complete telomere-to-telomere (T2T) assemblies have recently been generated for select individuals, high-quality, gapless Y chromosome references for East Asian lineages--specifically within well-characterized multi-omics reference cohorts--remain scarce. The Chinese Quartet (CQ), comprising monozygotic twin daughters and their parents, represents a premier national reference material system for genomic standardization. However, because the female offspring inherited their father's X chromosome rather than his Y chromosome, previous pedigree-based T2T assemblies (CQ v3.2) lacked a gapless Y chromosome sequence. Here, we present a complete, gapless T2T assembly of the Y chromosome (designated CQ_chrY) from the father of the CQ. This assembly was generated by integrating Oxford Nanopore Technologies ultra-long reads, PacBio high-fidelity reads, and Hi-C data, resulting in a sequence of 68.50 Mb. The assembly shows exceptional base accuracy (QV = 57.54) and structural assembly quality (CRAQ AQI = 99.781). We completely resolved the 40.09 Mb Yq12 heterochromatic region, a ~940 kb centromeric higher-order repeat array, and annotated 96 protein-coding gene models evaluated against GENCODE/T2T-Y standards, alongside 57.59 Mb (84.07%) of repetitive elements. CQ_chrY serves as a complementary paternal resource extending the CQ reference platform, providing a robust foundation for population genetics, structural variation analysis, and East Asian genomic standardization.

genomics↗

bioRxiv2026

MEIsensor: a deep-learning method for mobile element insertion discovery

Mobile element insertions (MEIs) are a critical source of structural variation in the human genome, yet their accurate detection remains challenging, particularly within repetitive regions and for structurally complex insertions. While long-read sequencing enables the direct recovery of inserted sequences, current analytical pipelines rely heavily on repeat-library alignment, which can limit both computational efficiency and accuracy. Here, we present MEIsensor, a deep learning framework designed to detect and classify MEIs directly from long-read sequencing data. Unlike conventional methodologies, MEIsensor performs direct, sequence-based annotation of Alu, LINE1, and SVA insertions. Evaluated against HGSVC benchmarks using HiFi long-read datasets, MEIsensor consistently outperformed existing tools across major MEI classes while substantially improving computational efficiency, demonstrating particularly pronounced gains for structurally complex SVA insertions. Furthermore, manual curation confirmed that MEIsensor successfully identifies MEIs hidden in highly repetitive genomic regions, including centromeric higher-order repeats, that were previously absent from benchmark datasets. Together, these advancements establish MEIsensor as an efficient framework for MEI analysis that promises to significantly advance our understanding of human genomic architecture and evolution.

bioinformatics↗

bioRxiv2026

PalmaClust: A graph-fusion framework leveraging the Palma ratio for robust ultra-rare cell type detection in scRNA-seq data

MotivationSingle-cell RNA sequencing (scRNA-seq) is routinely used to build atlases of tissues, resolve developmental trajectories, and characterize disease microenvironments. Yet many biologically and clinically meaningful populations--including transient progenitors, therapy-resistant tumor subclones, and antigen-specific lymphocytes--occur at very low frequencies (<1%) and are easily missed by standard clustering pipelines. Existing approaches often require extensive manual curation, rely on known marker genes, or trade sensitivity for unacceptable false positive rates due to the insensitivity of metrics like the Gini index to heavy-tailed distributions. A scalable, statistically grounded method is needed to sensitively detect rare populations while providing calibrated confidence and interpretable molecular signatures. ResultsWe present PalmaClust, a graph-fusion clustering framework that repurposes Palma ratio--a tail-sensitive inequality metric in sociology--to identify marker genes driven by extreme sparsity. PalmaClust constructs and fuses multiple K-Nearest Neighbor (KNN) graphs derived from complementary gene-selection statistics including the Palma ratio, Gini index, and Fano factor. It employs a local refinement strategy that re-prioritizes Palma-ranked genes within parent clusters. Benchmarking across diverse public scRNA-seq datasets confirms that PalmaClust consistently outperforms state-of-the-art baselines, improving rare-class F1 scores by at least 20% (absolute) while maintaining high global clustering stability. Further studies demonstrate that the Palma ratio-derived graph layer is essential for capturing ultra-rare signatures that other views miss. Availabilityhttps://github.com/wan-mlab/PalmaClust.

bioinformatics↗

bioRxiv2025

MetaPaCS: A novel meta-learning framework for pancreatic cancer subtype identification

As the third leading cause of cancer related deaths in the United States, pancreatic cancer (PaC) is a highly heterogenous malignancy that can be divided into a multitude of potential subtypes, with the main 4 consisting of aberrantly differentiated endocrine exocrine (ADEX), immunogenic, progenitor, and squamous. Each PaC subtype is characterized by unique molecular pathways and therapeutic characteristics. Identifying PaC molecular subtypes is essential for downstream patient risk stratification and tailored treatment design. Conventional wet-lab approaches for PaC subtyping like microdissection, histopathological studies or molecular profiling are often laborious, costly, and time-consuming. To address these concerns, we present MetaPaCS, a novel meta-learning framework to accurately identify PaC subtypes based on transcriptomics data only. Specifically, after preprocessing, the transcriptome-based feature vectors were classified by 10 base machine learning (ML) classifiers, whose prediction outputs were then combined with the initial preprocessed feature vectors to constitute a new set of ensemble feature vectors for a meta-learning model. Our meta-learning model could learn and leverage the diversity of different base classifiers to boost the prediction performance beyond any single ML model. Results based on 100 times ten-fold cross validation tests on benchmarking datasets demonstrated that MetaPaCS performed significantly better than existing state-of-the-art methods for PaC subtyping. In addition, our meta-learning model remarkably outperformed each individual base classifier, demonstrating that MetaPaCS could combine diverse results from multiple base classifiers to boost the ensemble performance. We believe that MetaPaCS is a promising tool for characterizing PaC subtypes and will have positive impacts on downstream risk stratification and personalized treatment design for PaC patients.

bioinformatics↗

bioRxiv2025

MOTLAB: A Weighted Multi-Omics Transfer Learning Approach to Mitigate Breast Cancer Racial Disparities

Breast cancer mortality remains markedly higher among Black women in the United States, while the limited representation of Black women in genomic datasets constrains the development of reliable clinical outcome prediction models for this racial group. Conventional transfer learning improves prediction in data-minority groups, but it does not simultaneously optimize omics-specific contributions in multi-omics integration and enhance learning from limited target-domain data. We developed MOTLAB to address both challenges through optimized multi-omics weighting, and data augmentation. In 1,085 female TCGA breast cancer cases, MOTLAB yielded the highest AUROC with three-omics integration across all progression-free interval time settings, reduced prediction error, and lowered calibration error in most settings compared with transfer learning alone. MOTLAB performance remained robust across molecular and clinical subgroups, while MOTLAB-derived risk scores captured breast cancer-relevant pathway, miRNA, and methylation signals. MOTLAB provides an integrated strategy that combines omics-specific weighting, transfer learning, and data augmentation to improve breast cancer clinical outcome prediction in data-minority groups.

bioinformatics↗

bioRxiv2025

A Multi-Modal Transfer Learning Framework to Reduce Health Disparities in Prostate Adenocarcinoma

Prostate cancer is the second most common cancer in men across the United States, of which prostate adenocarcinoma (PRAD) is the most common subtype. Despite remarkable progress has been made in PRAD diagnosis and prognosis, Black American men have been found to have disproportionately high incidence and mortality rate in PRAD compared with non-Hispanic White men in the United States. While machine learning (ML) methods like transfer learning (TL) have shown promise in reducing racial disparities in PRAD, its effectiveness is often compromised by the requirement for large-scale training datasets, which are challenging to obtain in clinical settings. In addition, existing ML approaches only leverage single-omics data without integrating multi-omics information. To address these concerns, we propose a novel approach called MOTLPRAD that develops a Multi-Omics integration Transfer Learning framework to reduce health disparities in PRAD. Specifically, we first investigated two multi-modal ensemble methods, Pearson correlation coefficient (PCC) based patient-pairwise similarity and variational autoencoder (VAE), to integrate different types of omics data. Then, we adopted a transfer learning model based on domain adaptation to pre-train the model on the majority group (e.g., non-Hispanic White Americans) and fine-tune the model using the minority group (e.g., Black Americans). To mitigate data imbalance across different ethnic groups, we leveraged the Synthetic Minority Oversampling Technique (SMOTE) to augment the sample size of minority groups, which could further improve the performance of reducing health disparities in PRAD. Results based on a series of multi-omics data (mRNA, miRNA, and methylation) from The Cancer Genome Atlas (TCGA) database suggested that our proposed model significantly outperformed conventional transfer learning models and other computational models for reducing racial disparities when predicting progression-free interval (PFI) prognosis for PRAD patients. In addition, our model also demonstrated that integrating multi-omics data could remarkably boost the performance of mitigating health disparities compared to using single-omics data. We believe our proposed framework could provide a new way to systematically mitigate health disparities towards underrepresented groups for PRAD diagnosis, prognosis, and treatment.

bioinformatics↗