Biology subjects

Jin, J.

Publications and source records attributed to Jin, J..

At least 55 records · Page 3Linked to original sources

bioRxiv2025

Melody: Decoding the Sequence Determinants of Locus-Specific DNA Methylation Across Human Tissues

DNA methylation is a fundamental epigenetic modification that plays crucial roles in transcriptional regulation, cellular differentiation, and genome stability. However, how locus-specific DNA methylation is determined by intrinsic DNA sequence remains poorly understood. Here, we introduce Melody, a deep learning framework that predicts DNA methylation from 10-kb genomic sequences, enabling the integration of both local and long-range sequence signals. Across 39 human tissues, Melody accurately predicts methylation profiles and consistently outperforms existing state-of- the-art methods in whole-chromosome, hypomethylated-region, and cell-type-specific benchmarks. Melody also generalizes to methylation quantitative trait locus (meQTL) effect prediction and identifies regulatory sequence motifs associated with methylation variability. To extend prediction beyond profiled tissues, we further develop Melody-G, which incorporates single-cell RNA-seq foundation model embeddings to infer methylation states in previously unseen cell types directly from transcriptomic data. Together, Melody provides a unified framework for linking genomic sequence and cellular state to DNA methylation and offers new insights into the regulatory logic governing the human methylome.

bioinformatics↗

bioRxiv2025

Tumor-on-Chip as a Personalised Platform for Rapid Drug-Testing in Breast Cancer

Breast cancer (BC) remains one of the most common malignancies worldwide and continues to pose major therapeutic challenges, emphasizing the need for functional models that can inform fast treatment selection. Currently, patient-derived xenografts (PDXs) and organoids (PDOs) are valuable biological models for functional precision oncology; however, variable success rates of establishment and prolonged timelines limit their clinical application for real-time drug testing. To overcome this, we developed a Tumor-on-Chip (ToC) platform that enables functional drug sensitivity profiling within a clinically actionable 4-day timeframe. We compared ToC models with PDX results, and demonstrated high reproducibility and strong concordance with in vivo PDX responses. Drug sensitivity was correctly identified ex vivo in 88% of PDX-responsive cases, while resistance was detected in 91%, with no false positives at clinically relevant drug concentrations. To facilitate clinical translation, we engineered a custom microfluidic chip optimized for minimal breast cancer biopsy material, yielding results similar to those obtained from resection samples. We ultimately demonstrated the proof-of-concept for applying this platform to patient samples as a further tool for guiding clinical decision-making and discriminated between resistance and sensitive profiles among patients. These findings demonstrate the feasibility and translational potential of ToC models for personalized drug profiling in breast cancer, laying the groundwork for their integration into real-time clinical decision-making workflows.

bioengineering↗

bioRxiv2025

Neutrophil extracellular traps-mediated thrombosis drive pyrrolizidine alkaloid-induced hepatic sinusoidal obstruction syndrome

Hepatic sinusoidal obstruction syndrome (HSOS) caused by pyrrolizidine alkaloid (PA) is a life-threatening disease with limited treatments, yet the initiating vascular event remains undefined. Neutrophil extracellular traps (NETs), as key driver of immunothrombosis in various thromboinflammatory diseases, may represent a critical yet uncharacterized component in the pathogenesis of PA-HSOS. Rats and mice were separately administered monocrotaline (MCT) or senecionine (SEN) via gavage to induce HSOS. Human hepatic sinusoidal endothelial cell (HHSEC) and primary human neutrophil treated with MCT, SEN, or corresponding metabolites were used as in vitro models. In separate experiments, animals received a neutrophil elastase (NE) inhibitor or neutrophil-depleting antibody to evaluate the contribution of NETs. We detected NETs within the sinusoids of necrotic liver lobes, as evidenced by robust immunostaining for NETs markers, accompanied with elevated level of circulating cell-free DNA, and hepatic citrullinated histone H3. Sinusoidal microthrombi containing VWF, CD41, and fibrin(ogen) were also stained positive for neutrophils in PA-HSOS models. Treatment with NE inhibitor or neutrophil depletion markedly attenuated PA-induced liver injury by concomitantly suppressing NETs formation and intrahepatic thrombogenesis. In vitro, conditioned medium from HHSEC exposed to MCT or SEN metabolites potently induced NETs formation. HHSEC injured by PA metabolites released CXCL8. Furthermore, hepatic mRNA and protein levels of CXCL1 and CXCL2 (murine homologous to human CXCL8) were up-regulated in PA-HSOS models, accompanied by a parallel elevation of their high-affinity receptors, CXCR1 and CXCR2. Our study identifies NETs-mediated immunothrombosis as a key driver in HSOS, which provides a promising target for PA-HSOS treatment.

pharmacology and toxicology↗

bioRxiv2025

Genome-wide SNVs revealed reticulate evolutionary history of an Eurasia-African songbird radiation

Phylogenomics has the power to uncover complex evolutionary hypothesis across taxa, thereby allowing important glimpses into the evolutionary history, such as incomplete lineage sorting (ILS) and introgression among closely related species. The Saxicola torquatus complex comprises several widely distributed insectivorous bird species across the Palearctic and Afrotropical realms. Nevertheless, the evolutionary history within this complex remains contentious, and previous findings suggest complication caused by cytonuclear discordance. Here, we conducted a phylogenomic analysis of nearly complete taxa within the complex using genome-wide SNVs to investigate potential causes of phylogenetic discordance across genomic regions and elucidate its evolutionary history. We detected multiple reticulate events during their diversification, leading to extensive phylogenetic discordance across the genome. The mitochondrial genome and W chromosome exhibited a consistent gene tree that conflicted with the species tree, with introgression being the most plausible explanation among several hypotheses. The Z chromosome was less affected by ILS/introgression and exhibited relatively stable local phylogenies compared to autosomes. Furthermore, we found that strongly introgressed regions were predominantly concentrated at the ends of smaller chromosomes, and the strength of introgression among species pairs showed consistently positive correlations, likely attributed to shared recombination hotspots in these regions. These findings stress the heterogeneous genomic landscape influenced by intricate evolutionary processes that endorse a reticulate evolution model for global speciation in the S. torquatus complex.

evolutionary biology↗

bioRxiv2025

YY1 relieves p300 autoinhibition to promote histone acetylation in advanced prostate cancer, enhancing the oncogenic signaling of Androgen Receptor Splice Variant 7

Genetic and epigenetic aberrations often act in concert to establish oncogenic transcriptomic programs in aggressive cancers. For example, the development of castration-resistant prostate cancer (CRPC), an advanced prostate cancer form, is closely associated with over-expression and/or hyper-activation of transcription factors (TFs) such as Androgen Receptor (AR) and Yin Yang 1 (YY1), as well as p300, a prominent histone acetyltransferase. How exactly these cancer-related lesions are coordinated to generate a malignant cell state remains elusive. Here, we demonstrate that YY1, which is frequently over-expressed in advanced prostate cancers, allosterically stimulates the acetyltransferase activity of p300 in cis, leading to the globally elevated acetylation of histone H3 lysine 18 and 27 (H3K18ac and H3K27ac). Mechanistically, YY1s N-terminal activation domain (AD) directly interacts with p300s TAZ2 domain, relieving the autoinhibition of p300 to facilitate substrate acetylation. Our integrated genome-wide mapping and transcriptomic studies reveal significant co-localization of genomic binding sites of YY1, androgen receptor splice variant 7 (AR-V7, a constitutively active form of AR) and p300 in CRPC cells, where the YY1-mediated p300 activation and resultant histone acetylation increases promote the oncogenic gene-expression programs downstream of YY1 and AR/AR-V7. Both in vitro and in vivo functional assays demonstrate a critical requirement of the above signaling for the advanced disease progression and drug resistance seen in CRPC. Altogether, this study uncovers that YY1 acts to alleviate p300s autoinhibition at target genes co-bound by oncogenic TFs (YY1 and/or AR/AR-V7) in CRPC, thereby sustaining tumorigenicity. Additionally, the blockade of YY1-mediated gene activation resensitizes CRPC to treatment to the clinic anti-AR agent (enzalutamide), which provides a rationale for overcoming the therapeutic resistance often seen in advanced prostate cancers.

cancer biology↗

bioRxiv2025

Accurate detection and quantification of single-base m6A RNA modification using nanopore signals with multi-view deep learning

N6-methyladenosine (m6A) is a crucial epitranscriptomic mark. While Nanopore Direct RNA Sequencing (DRS) enables transcriptome-wide detection, most existing methods neglect or lack the capacity to effectively process the intrinsically variable-length raw signals generated by DRS reads.Here, we present MultiNano, a multi-view deep learning framework that converts variable-length raw signals into image-like feature representations, effectively resolving the length inconsistency problem. By integrating raw signal and basecalling features, MultiNano enables accurate and comprehensive transcriptome-wide detection of m6A modifications. Our model achieved state-of-the-art (SOTA) performance in various tasks, including site-level prediction, read-level prediction, cross-species transfer learning, and modification rate estimation. Furthermore, the false positive control strategy implemented in MultiNano significantly enhances the models robustness and predictive accuracy, offering a powerful alternative to traditional thresholding-based filtering algorithms. Collectively, our approach provides novel insights for the absolute quantification and single-base resolution of RNA modifications.

bioinformatics↗

bioRxiv2025

Lysosome-like structures in the nucleolus mediate H2Aub degradation under glucose starvation

Glucose starvation leads to the degradation of intracellular proteins to generate amino acids for energy production. However, it remains unclear which proteins undergo rapid degradation to supply amino acids as energy sources during glucose starvation. To investigate the molecular mechanism underlying the transition of carbon source from glucose to proteins under glucose starvation, we first employed quantitative proteomics to examine the impact of glucose deprivation on protein levels. We discovered that histone degradation produces the highest amount of glutamine during glucose starvation. Further investigation revealed that histone H2A undergo degradation in the nucleolus in the form of mono-ubiquitination (H2Aub). The nucleolus contains lysosome-like structures that degrade monoubiquitinated histones, potentially participating in the carbon source shift from glucose to amino acids for energy production. Through preliminary screening, we identified Midnolin as the receptor for H2Aub, mediating its degradation in nucleolar lysosome-like structures. Dysregulation of carbon source transition from glucose to proteins, through knockdown of Midnolin results in cell cycle arrest, affecting cellular growth and proliferation. In conclusion, our findings demonstrate that nucleolar lysosome-like structures degrading monoubiquitinated histones exist during glucose starvation, which may play crucial roles in the carbon source transition from glucose to proteins, providing new potential therapeutic targets for metabolic disorders.

cell biology↗

bioRxiv2025

FBL as a carrier mediate H2Aub degradation in the lysosome-like structures

Under glucose starvation, mammalian cells form lysosome-like structures within the nucleoli to mediate the metabolic shift of carbon sources from glucose to amino acids generated through H2Aub degradation for energy supply, but the mechanism by which lysosome-like structures mediate the degradation of monoubiquitinated histones during glucose starvation remains unclear. Through screening, we discovered that the fibrillarin (FBL) protein in the nucleolus acts as a carrier for mono-ubquitinated histone H2A (H2Aub), regulating the balance between BMI1-mediated mono-ubiquitination of H2A and USP36-mediated deubiquitination of H2Aub. Under glucose starvation, FBL undergoes aggregation, leading to a significant reduction in its interaction with BMI1, accompanied by markedly decreased binding to histone H2A and H2Aub. Concurrently, Midnolin, which serves as a receptor for monoubiquitinated histones, exhibits increased interaction with FBL during glucose starvation. USP36, FBL, Midnolin, and BMI1 collectively form a complex responsible for degrading monoubiquitinated histones. Furthermore, knockdown of FBL, BMI1, or USP36 resulted in cell cycle arrest at S phase and severely compromised cell viability. In summary, we identified a lysosome-like structure in the nucleolus that mediates the degradation of H2Aub. This structure plays a critical role in regulating cell cycle progression and survival, and may offer novel potential therapeutic targets for cancer.

cell biology↗

bioRxiv2025

Deep-learning-enabled multi-omics analyses for prediction of future metastasis in cancer

Metastasis remains the leading cause of cancer-related mortality, yet predicting future metastasis is a major clinical challenge due to the lack of validated biomarkers and effective assessment methods. Here, we present EmitGCL, a deep-learning framework that accurately predicts future metastasis and its corresponding biomarkers. Based on a comprehensive benchmarking comparison, EmitGCL outperformed other computational tools across six cancer types from seven cohorts of patients with superior sensitivity and specificity. It captured occult metastatic cells in a patient with a lymph node-negative breast cancer, who was declared to have no evidence of disease by conventional imaging methods but was later confirmed to have a metastatic disease. Notably, EmitGCL identified HSP90AA1 and HSP90AB1 as predictable biomarkers for future breast cancer metastasis, which was validated across five independent cohorts of patients (n=420). Furthermore, we demonstrated YY1 transcription factor as a key driver of breast cancer metastasis which was validated through in-silico and CRISPR-based migration assays, suggesting that YY1 is a potential therapeutic target for deterring metastasis.

bioinformatics↗

bioRxiv2025

Grapevine phyllosphere pan-metagenomics reveals pan-microbiome structure, diversity, and functional roles in downy mildew resistance

Grapevines are among the most economically important fruit crops, and microbiome profoundly influence their health, yield, and quality. However, mechanistic insights into microbiome-orchestrated grapevine biology remain limited. Here, we conduct large-scale pan-metagenomic and pan-metatranscriptomic analyses of the phyllosphere microbiome from 107 grapevine accessions spanning 34 Vitis species. We show that the grapevine core microbiome is dominated by phylum Bacillota and Pseudomonadota. Leveraging PacBio sequencing, we assembled 19 high-quality metagenome-assembled genomes (MAGs) from the grapevine pan-microbiome, representing the first MAG reconstruction in plant-associated microbial communities using PacBio reads. These MAGs encode genes associated with antibiotic resistance, secondary metabolism, and carbohydrate-active enzymes (CAZymes), which could potentially influence grapevine biology. During downy mildew (DM) infection, DM-resistant grapevines exhibit significantly higher microbial network complexity than susceptible counterparts. Among the key taxa contributing to this complexity, Bacillota emerged as the dominant phylum with displaying strong abundance correlations with phylum Euglenozoa and Cyanobacteriota, and an isolated Bacillota species from the grapevine leaves, Bacillus cereus, demonstrated potent biocontrol activity against DM infection. Pan-metatranscriptomic analysis further revealed significant upregulation of eukaryotic microbial genes involved in primary and secondary metabolism. This pan-metagenomic study offers unprecedented insights into the complex structure, diversity and functional roles of grapevine phyllosphere microbiome, and presents valuable genomic and microbial resources for microbiome research and engineering to enhance viticulture productivity and quality.

microbiology↗

bioRxiv2025

Characterization of MEK1/2 Degraders Uncovers a Kinase-Independent Role for MEK1/2 in the Stabilization and Maturation of CRAF

Altered MAPK signaling frequently occurs in human disease. MEK1 and MEK2 (MEK1/2) are central protein kinases in the MAPK signaling cascade that phosphorylate ERK1/2 promoting cell growth. MEK1/2 degraders offer a strategy to characterize both kinase-dependent and independent functions of MEK1/2. Here, we discovered that MEK1/2 degradation, but not kinase inhibition, caused the subsequent degradation of upstream kinase CRAF via a cell-intrinsic mechanism. MEK1/2 binding to CRAF, but not MEK1/2 catalytic activity, was required for CRAF protein stability and maturation to a functional kinase. In the absence of MEK1/2, a minor pool of newly synthesized immature CRAF that had anti-apoptotic functions remained. Finally, we showed that a stable primed CRAF-MEK1/2 signaling complex existed in cells that required RAS binding to potentiate MEK-ERK phosphorylation. Together, weve discovered a previously unrecognized kinase-independent function of MEK1/2, while contextualizing MEK1/2 as an integral component of the CRAF activation cycle beyond the conventional CRAF-MEK kinase- substrate paradigm.

cancer biology↗

bioRxiv2025

HGCPep: A Hypergraph-based Deep Learning Model for Enhancing Representation of Peptide Features in Cancer-associated ncPEPs Identification

The emergence of non-coding RNA-encoded small peptides (ncPEPs) has sparked significant interest within the realm of cancer immunotherapy, owing to their potential as valuable therapeutic targets and biomarkers. The identification and characterization of cancer-associated ncPEPs assume a crucial role in propelling cancer research forward and augmenting our comprehension of immune-related processes. However, prevailing methods for cancer-associated ncPEPs identification primarily rely on sequence order, neglecting the latent relationships between peptides. As widely acknowledged in the central dogma, peptides are translated from RNA via a many-to-many mapping. In this study, we capitalize on the translation relationship and introduce a hypergraph-based model called HGCPep to enhance the representation of peptide features. Specifically, RNA is regarded as a hyperedge connecting the peptides that are translated from it. Our experimental results validate that the HGCPep approach, leveraging the strengths of hypergraph and convolutional neural networks, outperforms alternative methodologies in the recognition of ncPEPs. Additionally, employing a reduction tool, we visualize the results of multi-label classification within a 2D latent space, shedding light on how multi-label classification task influence the representation of peptides. The dataset and source code of our proposed method can be found via https://github.com/Longwt123/HGCPep_Github.

bioinformatics↗

bioRxiv2025

Transcriptional control of T cell tissue adaptation and effector function in infants and adults

The first years of life are essential for the development of memory T cells, which rapidly populate the bodys diverse tissue sites during infancy. However, the degree to which tissue memory T cell responses in early life reflect those during adulthood is unclear. Here, we use single cell RNA-sequencing of resting and ex vivo activated T cells from lymphoid and mucosal tissues of infant (aged 2-9 months) and adult (aged 40-65 years) human organ donors to dissect the transcriptional programming of memory T cells over age. Infant memory T cells demonstrate a unique stem-like transcriptional profile and tissue adaptation program, yet exhibit reduced activation capacity and effector function relative to adults. Using CRISPR-Cas9 knockdown, we define Helios (IKZF2) as a critical transcriptional regulator of the infant-specific tissue adaptation program and restricted effector state. Our findings reveal key transcriptional mechanisms that control tissue T cell fate and function in early life.

immunology↗

bioRxiv2025

Loss of tumor cell MHC Class II drives insensitivity of BRAF-mutant anaplastic thyroid cancers to MAPK inhibitors

Cancer cells present neoantigens dominantly through MHC class I (MHCI) to drive tumor rejection through cytotoxic CD8+ T-cells. There is growing recognition that a subset of tumors express MHC class II (MHCII), causing recognition of antigens by TCRs of CD4+ T-cells that contribute to the anti-tumor response. We find that mouse BrafV600E-driven anaplastic thyroid cancers (ATC) respond markedly to the RAF + MEK inhibitors dabrafenib and trametinib (dab/tram) and that this is associated with upregulation of MhcII in cancer cells and increased CD4+ T-cell infiltration. A subset of recurrent tumors lose MhcII expression due to silencing of Ciita, the master transcriptional regulator of MhcII, despite preserved interferon gamma signal transduction, which can be rescued by EZH2 inhibition. Orthotopically-implanted Ciita-/- and H2-Ab1-/- ATC cells into immune competent mice become unresponsive to the MAPK inhibitors. Moreover, depletion of CD4+, but not CD8+ T-cells, also abrogates response to dab/tram. These findings implicate MHCII-driven CD4+ T cell activation as a key determinant of the response of Braf-mutant ATCs to MAPK inhibition.

cancer biology↗

bioRxiv2025

CFTR High Expresser BEST4+ cells are pH-sensing neuropod cells: new implications for intestinal physiology and Cystic Fibrosis disease

Single-cell RNA sequencing (scRNA-seq) studies identified a novel subpopulation of epithelial cells along the rostrocaudal axis of human intestine specifically marked by bestrophin 4 (BEST4) that are enriched for genes regulating pH, GPCR acid-sensing receptors, satiety, cGMP signaling, HCO3- secretion, ion transport, neuropeptides, and paracrine hormones. Interestingly, BEST4+ cells in the proximal small intestine express CFTR but have not been linked to the previously described CFTR High Expresser Cell (CHE) subpopulation in rat and human intestine. ScRNA-seq studies in rat jejunum identified CHEs and a gene expression profile consistent with human small intestinal BEST4+ and neuropod cells. Protein immunolocalization confirmed that CHEs express CFTR, BEST4, neuropod proteins, high levels of intracellular uroguanylin (UGN), guanylyl cyclase-C (GC-C), and the proton channel otopetrin 2 (OTOP2), and display long basal processes connecting to neurons. OTOP2, GC-C, and CFTR traffic robustly into the apical domain of CHEs in response to acidic luminal conditions, indicating their roles in luminal pH regulation. In the {Delta}F508 cystic fibrosis (CF) rat jejunum, the loss of apical CFTR did not affect BEST4 protein expression in CHEs. However, there was an increased abundance of CHE cells in the {Delta}F508 rat jejunum compared to wild-type animals. Furthermore, {Delta}F508 rat CHEs expressed higher levels of GC-C at the apical domain compared to wild-type. These data implicate CHEs in intestinal CF disease pathogenesis. NEW & NOTEWORTHYThis is the first study to identify CFTR High Expresser cells in the rat small intestine as neuropod cells capable of sensing and responding to luminal pH. This study also provides the first characterization of CFTR and relevant mRNA and proteins in CHEs in CF rat models that provide insights into the significance of CHEs to CF intestinal disease.

physiology↗

bioRxiv2025

Fibroblast contractility drives network reorganization and epithelial proliferation in intestinal polyposis

Fibroblasts are critical regulators of epithelial homeostasis through mechanics and signaling. However, the regulatory principles governing fibroblast behaviors are largely unknown. Fibroblast dysregulation has emerged as a pathological contributor in epithelial diseases such as intestinal polyposis. Here, using an inducible Bmp-loss-of-function polyposis model, we define stepwise mechanisms unraveling how dysregulated signaling perturbs fibroblast behaviors, which in turn, disrupt their function to regulate epithelial homeostasis. Intriguingly, the first initiating event leading to epithelial polyps was architectural and impacted the fibroblasts, not the epithelium. Bmp signaling inhibition caused fibroblasts to become hypercontractile, leading to their reorganization from an elaborate network into collapsed clusters beneath crypts. Disrupted fibroblast mechanics and compartmentalization preceded epithelial hyperproliferation in polyps. Using in vitro models, we show that fibroblast hypercontractility not only disrupted fibroblast organization, but also enhanced the ability of fibroblasts to support epithelial growth, leading to epithelial dysregulation. Overall, our studies reveal stepwise regulatory mechanisms underlying fibroblast signaling, mechanics and organization critical for their function to regulate epithelial homeostasis.

cell biology↗

bioRxiv2024

PRMT1-Mediated Metabolic Reprogramming Promotes Leukemogenesis

Copious expression of protein arginine methyltransferase 1 (PRMT1) is associated with poor survival in many types of cancers, including acute myeloid leukemia. We observed that a specific acute megakaryocytic leukemia (AMKL) cell line (6133) derived from RBM15-MKL1 knock-in mice exhibited heterogeneity in Prmt1 expression levels. Interestingly, only a subpopulation of 6133 cells expressing high levels of Prmt1 caused leukemia when transplanted into congenic mice. The PRMT1 inhibitor, MS023, effectively cured this PRMT1-driven leukemia. Seahorse analysis revealed that PRMT1 increased the extracellular acidification rate (ECAR) and decreased the oxygen consumption rate (OCR). Consistently, PRMT1 accelerated glucose consumption and led to the accumulation of lactic acid in the leukemia cells. The metabolomic analysis supported that PRMT1 stimulated the intracellular accumulation of lipids, which was further validated by FACS analysis with BODIPY 493/503. In line with fatty acid accumulation, PRMT1 downregulated the protein level of CPT1A, which is involved in the rate-limiting step of fatty acid oxidation. Furthermore, administering the glucose analogue 2-deoxy-glucose (2-DG) delayed AMKL progression and promoted cell differentiation. Ectopic expression of Cpt1a rescued the proliferation of 6133 cells ectopically expressing PRMT1 in the glucose-minus medium. In conclusion, PRMT1 upregulates glycolysis and downregulates fatty acid oxidation to enhance the proliferation capability of AMKL cells. Graphic Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=83 SRC="FIGDIR/small/628174v2_ufig1.gif" ALT="Figure 1"> View larger version (26K): org.highwire.dtl.DTLVardef@1ae96borg.highwire.dtl.DTLVardef@d27916org.highwire.dtl.DTLVardef@ba63corg.highwire.dtl.DTLVardef@cee9af_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2024

Identifying Effect Modification of Latent Population Characteristics on Risk Factors with a Sparse Varying Coefficient Regression

Leveraging observational data to understand the associations between risk factors and disease outcomes and conduct disease risk prediction is a common task in epidemiology. While traditional linear regression and other machine learning models have been extensively implemented for this task, the associations between risk factors and disease outcomes are typically deemed fixed. In many cases, however, such associations may vary by some underlying features of the individuals, which may involve certain subpopulation characteristics and environmental factors. While data for these latent features may not be available, the observed data on risk factors may have captured some proportion of the variation in these features. Thus extracting latent factors from risk factors and incorporating this effect modification into the model may better capture the underlying data structure and improve inference. We develop a novel regression model with some coefficients varying as functions of latent features extracted from the risk factors. We have demonstrated the superiority of our approach in various data settings via simulation studies. An application on a dataset for lung cancer patients from The Cancer Genome Atlas (TCGA) Program showed that our approach led to a 6% - 118% increase in (AUC-0.5) for distinguishing between different lung cancer stages compared to the classic lasso and elastic net regressions and identified interesting latent effect modifications associated with certain gene pathways.

genetics↗