Biology subjects

Wen, J.

Publications and source records attributed to Wen, J..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

Integrative genomic insights into parallel local adaptation to Danxia and Karst edaphic islands in the endangered tree Firmiana danxiaensis

Uncovering the genetic basis of local adaptation is a central goal in evolutionary biology. Edaphic islands, such as Karst (KL) and Danxia (DL) landforms, provide ideal natural laboratories for this endeavor. However, the mechanisms underlying parallel adaptation to contrasting edaphic extremes within a single species remain poorly understood. Firmiana danxiaensis, an endangered tree endemic to both KL and DL in the Nanling Mountains, offers a rare opportunity to address this question. We integrated a chromosome-level genome assembly, population genomic analysis of 225 individuals from 20 populations across its entire range, and genotype-environment association (GEA) with high-resolution soil and climate data. The relatively large genome of F. danxiaensis (1.51 Gb) reveals two ancient whole-genome duplications and a Quaternary burst of LTR retrotransposons, which, together with significant expanded drought-related transcription factor (TF) families, together constitute its genomic basis of adaptation. Population structure resolved four genetic lineages with distinct demography and inbreeding patterns. GEA identified stronger local adaptation in DL than KL, driven mainly by soil factors (e.g., cadmium, zinc), while climate variables (e.g., Bio9, Bio11) predominated in KL. Despite these divergent drivers, drought stress and elevated metal ion acted as major selective pressures in both landforms. Furthermore, six key TFs (e.g., NAC090, bHLH42, MAPKKK20) under both local adaptation and positive selection in DL are involved in drought response and metal ion binding, facilitating survival in stressful habitats. Our findings provide novel insights into parallel local adaptation to contrasting edaphic islands and offer concrete genomic guidance for the conservation of F. danxiaensis. Significance statementThe genetic basis of parallel local adaptation to contrasting environments remains unclear. Using the endangered tree Firmiana danxiaensis, we uncover genomic mechanisms enabling its adaptation to both Karst and Danxia edaphic islands. We demonstrate that adaptation is primarily driven by soil factors in Danxia and climate in Karst, generating divergent genomic signatures yet converge on drought response. These findings establish a framework for understanding parallel local adaptation and offer concrete genomic guidance for conservation.

evolutionary biology↗

bioRxiv2025

Delivery of collagen-encoding mRNA to the skin for cosmetic applications

Collagens, the predominant structural proteins in mammalian, maintain tissue integrity through its unique triple-helical conformation. Traditionally, cosmetic collagens have been administered as hydrolyzed peptides derived from animal sources, which exhibit poor bioavailability and functional efficacy due to limited skin penetration and disrupted tertiary/quaternary structures. To overcome these limitations, we developed a novel system for in vivo collagen expression. This system utilizes the topical application of in vitro synthesized, AI-designed mRNAs encoding types I, III, and XVII collagens to improve human skin condition. Significant upregulation of target collagen expression was observed in dermal fibroblasts, leading to improved skin condition through in situ assembly of collagens in natural conformation.

bioengineering↗

bioRxiv2025

NucEL: Single-Nucleotide ELECTRA-Style Genomic Pre-training for Efficient and Interpretable Representations

Pre-training large language models on genomic sequences has become a powerful approach for learning biologically meaningful representations. While masked language modeling (MLM)-based approaches, such as DNABERT and Nucleotide Transformer (NT), achieve strong performance, they are hindered by inefficiencies due to partial token supervision, pre-training/fine-tuning mismatches, and high computational costs. We introduce NucEL, the first ELECTRA-style pre-training framework for genomic foundation models, which overcomes these challenges. Through a discriminator network identifying tokens modified by a generator, Nu-cEL achieves comprehensive token-level supervision across all sequence positions, thereby markedly improving training efficiency relative to the partial supervision of masked positions inherent in MLM frameworks. By integrating ModernBERTs architectural advancements, including hybrid local-global attention and flash attention mechanisms, NucEL establishes an optimized BERT architecture for genomic sequence modeling. Unlike traditional methods that tokenize genomic sequences into 6-mers, NucEL implements single-nucleotide tokenization, enabling fine-grained resolution and improving both efficiency and interpretability. Pre-trained on the human genome only, NucEL achieves state-of-the-art performance on benchmark datasets across diverse downstream tasks in both human and non-human species, including regulatory element identification (e.g., promoters, enhancers), transcription factor binding prediction in human and mouse, open chromatin region classification, and histone modification profiles, surpassing MLM-based models of similar size and rivaling models 25 times larger, such as NT. Ablation studies provide critical insights into tokenization and masking strategies, optimizing ELECTRA-style pretraining for DNA sequences. Attention analyses reveal NucELs superior ability to capture biologically relevant sequence motifs compared to NT, offering valuable insights into its hierarchical learning process and regulatory element modeling capabilities. This work highlights the potential of ELECTRA-style pretraining as an efficient and effective strategy for advancing genomic representation learning with broad implications for future genomic research.

genomics↗

bioRxiv2025

MuSHIN: A multi-way SMILES-based hypergraph inference network for metabolic model reconstruction

Genome-scale metabolic models (GEMs) are indispensable tools for probing cellular metabolism, enabling predictions of metabolic fluxes, guiding strain optimization, and advancing biomedical research. However, their predictive capacity is often compromised by incomplete reaction networks, stemming from gaps in biochemical knowledge, annotation inaccuracies, and insufficient experimental validations. Here we present MuSHIN (Multi-way SMILES-based Hypergraph Interface Network), a novel deep hypergraph learning method that integrates network topology with biochemical domain knowledge to predict missing reactions in GEMs. Evaluated on 926 high- and intermediate-quality GEMs with artificially removed reactions, MuSHIN significantly outperforms state-of-the-art methods, achieving up to a 17% improvement across multiple metrics and maintaining robust recovery even under severe network sparsity. Furthermore, MuSHIN substantially enhances phenotypic predictions in 24 draft GEMs associated with fermentation by resolving critical metabolic gaps, as validated against experimental measurements. Together, these findings highlight MuSHINs potential to advance GEM reconstruction and accelerate discoveries in systems biology, metabolic engineering, and precision medicine.

bioinformatics↗

bioRxiv2025

PathoEye: a deep learning framework for the whole-slide image analysis of skin tissue

ObjectiveTo provide an interpretable computational framework for the examination of whole-slide images (WSI) in skin biopsies, PathoEye focuses on the dermis-epidermis junctional (DEJ) areas, also known as the basement membrane zone (BMZ), to enrich the pathological features of various skin diseases. MethodsWe presented PathoEye for WSI analysis in dermatology, which integrates epidermis-guided sampling, deep learning and radiomics. It enables unsupervised semantic segmentation of the BMZ and extracts distinct features associated with various skin conditions. ResultsPathoEye performs comparably with existing methods in binary classification tasks, while outperforming them in multi-classification tasks involving different skin conditions. It enables the investigation of histopathological aberrations in aged skin compared with young skin. Additionally, it highlighted the texture changes in the BMZ of young skin compared with aged skin. Further experimental analyses revealed that senescence cells were enriched in the BMZ, and the turnover of basement membrane (BM) components, including COL17A1, COL4A2, and ITGA6, was increased in aged skin. ConclusionPathoEye is a comprehensive tool for characterizing the unique features of different skin conditions associated with BMZ, thereby ensuring its potential applications in skin disease diagnosis and treatment planning.

bioinformatics↗

bioRxiv2025

Treatment with the ribosome biogenesis inhibitor CX-5461 increases platelet count in humans and enhances murine megakaryopoiesis

Thrombocytopenia is a common and serious complication of anticancer therapies. Here, we identify a novel thrombopoietic activity of the first-in-class ribosome biogenesis inhibitor CX-5461. In a phase I trial, 56% (9/16) of patients exhibited up to a 34% increase in platelet count following a single dose of CX-5461. In mice, CX-5461 elicited a rapid, reversible, and sustained [~]1.7-fold increase in platelet numbers without altering platelet function, lifespan, or inflammatory cytokines. Bone marrow analysis revealed a specific expansion of megakaryocytes (MKs), increased Sca1 MKs, and selective enrichment of MK-biased multipotent progenitor 2, independent of thrombopoietin (TPO) or c-mpl signalling. CX-5461 also mitigated carboplatin-induced thrombocytopenia, accelerating platelet recovery. Single-cell RNA sequencing and RNA velocity analysis confirmed enhanced differentiation of MK progenitors. These findings demonstrate that inhibition of ribosome biogenesis promotes TPO-independent megakaryopoiesis and identifies a previously unrecognised therapeutic opportunity to support platelet recovery in cancer treatment and potentially other thrombocytopenic states.

cell biology↗

bioRxiv2025

Synthetic essentiality of TRAIL/TNFSF10 in VHL-deficient renal cell carcinoma

Clear cell renal cell carcinoma (ccRCC) is the most common and aggressive subtype of kidney cancer. Loss of von Hippel-Lindau (VHL) and the consequent activation of hypoxia-inducible factor- (HIF, especially HIF2) plays an essential role in ccRCC initiation and progression. The approved HIF2 inhibitor belzutifan faces the challenge of resistance, presenting an opportunity of co-targeting HIF2 and another vulnerability. This study elucidates the synthetic essentiality of TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) in VHL-deficient ccRCC, uncovering a novel reciprocal regulation between HIF2 and TRAIL. TRAIL was identified as a direct transcriptional target of HIF2 and paradoxically found to be crucial for cell proliferation, primarily by activating the p38 MAPK pathway and facilitating G1/S phase transition. Depletion of endogenous TRAIL or inhibition of HIF2 with belzutifan sensitizes ccRCC cells to recombinant TRAIL, presenting a promising avenue for combination therapy to overcome both TRAIL resistance and belzutifan resistance in treating ccRCC. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=162 SRC="FIGDIR/small/621197v1_ufig1.gif" ALT="Figure 1"> View larger version (36K): org.highwire.dtl.DTLVardef@1d603b9org.highwire.dtl.DTLVardef@18e6dc7org.highwire.dtl.DTLVardef@1c953cdorg.highwire.dtl.DTLVardef@9cd144_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2025

CS-Fold: Advancing RNA Structure Predictions through Phylogenetic Modelling of Compensatory Mutations in Deep Neural Networks

Accurate prediction of RNA secondary structures is essential for understanding the conformation, function, and interactions of RNA. Leveraging co-evolutionary information across species through multiple sequence alignments (MSAs) has been proven to be effective in improving molecular structure predictions. However, existing deep learning approaches do not explicitly incorporate compensatory substitutions along the phylogenetic trees, which are crucial for capturing structural conservation through evolution. To address this, we developed CS-Fold, a novel deep learning framework that integrates compensatory substitutions likelihoods as constraints using likelihood estimation and Monte Carlo algorithms. These likelihoods, representing evolutionary changes along phylogenetic trees, are encoded in a sparse matrix that guides the attention mechanism of a Pairformer-based architecture. A custom loss function and an unrolled post-processing algorithm enforce adherence to the solution space constrained by these evolutionary constraints. CS-Fold achieves a substantial 5% improvement in the F1 score compared to the current mainstream approaches, demonstrated through evaluations on cross-family datasets, including 604 human RNA families from the Rfam database. Our model offers novel insights and incorporates additional evolutionary information beyond traditional MSAs and folding strategies, providing a robust and innovative solution for RNA secondary structure prediction.

bioinformatics↗

bioRxiv2025

A Novel Glia-immune humanized mouse model for investigation of HIV CNS infection and neuroinflammation.

Dysregulation in neuroimmune interactions drives pathology in diverse neurological disorders. However, human cell-based in vivo models are limited for clinically relevant mechanistic studies and drug testing. To address this, we developed a novel humanized mouse model integrating all four major types of human glia - astrocytes, oligodendrocyte precursor cells (OPCs), oligodendrocytes, and microglia - in the brain, alongside donor-matched human immune system in peripheral blood and lymphoid tissues. We applied this model to study HIV-associated neurocognitive disorders, which can persist in tissue reservoirs including in the brain despite viral suppression by antiretroviral therapy. These glia-immune humanized mice supported robust HIV-1 replication in the peripheral blood, lymphoid tissues, and the brain, and recapitulated the synapse loss observed in patients. Furthermore, HIV-infected mice exhibited heightened inflammation in both peripheral and brain tissues. Notably, OPCs and oligodendrocytes in HIV-infected brains adopted an immune activated phenotype, upregulating interferon-stimulated genes, highlighting understudied roles of oligodendroglia in HIV infection. RNA-sequencing of human glia identified the upregulation of inflammasome-associated genes and the downregulation of genes involved in transcriptional, epigenetic, and metabolic regulation. Overall, this model provides a powerful in vivo platform to investigate human neuroimmune interactions relevant to diverse neurological disorders, offering critical insights into HIV CNS infection, pathology, and potential therapeutic strategies targeting CNS reservoirs and neuroinflammatory pathways.

immunology↗

bioRxiv2025

Mutations adjacent to the nucleotide-binding cleft of Arabidopsis thaliana ACTIN7 confer resistance to the actin-disrupting compound latrunculin B

A forward genetic screen identified a dominant-negative Arabidopsis thaliana mutant resistant to growth inhibition caused by the actin-disrupting compound latrunculin B (LatB). Map-based cloning combined with whole-genome sequencing revealed that the mutant referred to here as lbr1 for LatB resistant1 had a point mutation in the AT5G09810 gene, which encodes the vegetative actin (ACT) isoform ACT7. The cytosine to thymine mutation in the second exon of ACT7 of lbr1 led to substitution of proline to serine at position 34 (P34S) adjacent to the nucleotide-binding cleft of the ACT7 protein. Confirmation that ACT7 is the causal gene for the lbr1 phenotype was achieved through transgenic complementation with ACT7 wild type (ACT7WT) and ACT7 P34S (ACT7P34S) constructs. ACT7P34 also rescued the seedling developmental defects and conferred partial resistance to LatB in the recessive act7-5 mutant. Furthermore, expressing a P34S mutation in ACT2 (ACT2P34S), another vegetative ACT isoform, conferred partial LatB resistance to wild type. Finally, site-directed mutagenesis of ACT7 amino acid residues forming putative hydrogen bonds with LatB, based on yeast and mammalian actin docking and structural analyses, reveals domains adjacent to the actin nucleotide-binding cleft crucial for LatBs effects on the plant actin cytoskeleton. HighlightCharacterization of the dominant-negative lbr1 mutant uncovers amino acid residues in the actin protein crucial for latrunculins mechanism of action in plants.

cell biology↗

bioRxiv2025

Variational Autoencoder-based Model Improves Polygenic Prediction in Blood Cell Traits

Genetic prediction of complex traits, enabled by large-scale genomic studies, has created new measures to understand individual genetic predisposition. Polygenic Risk Scores (PRS) offer a way to aggregate information across the genome, enabling personalized risk prediction for complex traits and diseases. However, conventional PRS calculation methods that rely on linear models are limited in their ability to capture complex patterns and interaction effects in high-dimensional genomic data. In this study, we seek to improve the predictive power of PRS through applying advanced deep learning techniques. We show that the Variational AutoEncoder-based model for PRS construction (VAE-PRS) outperforms currently state-of-the-art methods for biobank-level data in 14 out of 16 blood cell traits, while being computationally efficient. Through comprehensive experiments, we found that the VAE-PRS model offers the ability to capture interaction effects in high-dimensional data and shows robust performance across different pre-screened variant sets. Furthermore, VAE-PRS is easily interpretable via assessing the contribution of each individual marker to the final prediction score through the SHapley Additive exPlanations (SHAP) method, providing potential new insights in identifying trait-associated genetic variants. In summary, VAE-PRS presents a novel measure to genetic risk prediction by harnessing the power of deep learning methods, which could further facilitate the development of personalized medicine and genetic research.

bioinformatics↗

bioRxiv2025

Intestinal enteroendocrine cell subtype differentiation and hormone production in zebrafish

Enteroendocrine cells (EECs) are rare sensory cells in the intestinal epithelium that coordinate digestive physiology by secreting a diverse repertoire of peptide hormones. These hormones are the main effectors of EEC function, and their characterization requires direct observation by mass spectrometry due to the specialized protein cleavage and posttranslational modifications that yield their mature forms. Based on the distinct subset of hormones they predominantly secrete, EECs can be categorized into subtypes. How each EEC subtype is specified, however, remains poorly understood. Here we describe EEC subtype differentiation and hormone production in the zebrafish. Using single-cell RNA sequencing data, we identified EEC progenitors and six EEC subtypes in zebrafish and revealed that their expression profiles are consistent across larval and adult stages. Mass spectrometry analysis of isolated zebrafish EECs identified highly processed peptides derived from 18 of 21 hormone coding genes expressed by EECs, yielding a catalog of >400 unique EEC hormone peptides. We assembled reporters for zebrafish EEC subtypes to test the lineage relationships between EEC subtypes and the EEC progenitor population, which expresses neurogenin3. Despite its essential role in mammalian EEC differentiation, we found that selective cytotoxic ablation of neurogenin3+ cells in zebrafish only reduced a subset of EEC subtypes. Finally, we discovered that selective ablation of ghrelin+ EECs reduced a different subset of EEC subtypes, together suggesting that neurogenin3+ and ghrelin+ cells serve as distinct precursors for separate EEC subtypes. We anticipate these observations and resources will facilitate future studies in the zebrafish to discern the developmental biology, physiology, and endocrinology of EEC subtypes.

developmental biology↗

bioRxiv2025

Extreme high-diversity barcoded HIV library enables single cell multi-omics analysis of viral expression and fate determination

HIV-1 persists as a global health challenge due to its ability to establish latent reservoirs that evade eradication by antiretroviral therapy (ART). This study introduces a high-diversity barcoded HIV-1 library combined with a novel peptide barcode system to enable single-cell multi-dimensional analysis of viral integration, transcription, splicing, and translation of distinct viral mRNA isoform. Our method revealed over 80 distinct HIV-1 splicing isoforms and uncovered significant temporal variation in mRNA diversity, driven by integration site and chromatin context. By linking HIV-1 5UTR with peptide barcoding, we quantified the translation efficiency of alternative 5' UTRs and demonstrated correlations with abundance of each viral mRNA isoform. Additionally, we showed that latency reversal agents, such as SAHA and Bryostatin, selectively reactivate proviruses depending on their genomic and chromatin context and leads to distinct viral RNA splicing program. This platform revealed the diversity of HIV-infected cells at levels of transcripts as well as at proteins, and subsequently the fate determination. The detailed understanding of the stochastic diversity and dynamic nature of HIV-1 latency, inform the development of targeted therapies to eliminate latent reservoirs.

microbiology↗

bioRxiv2024

Target enrichment phylogenomics and biogeographic analyses unravel rapid radiation and reticulate evolution between Hainan-South China mainland -Vietnam in Section Nintooa (Lonicera, Caprifoliaceae)

The South China mainland and nearby islands are biodiversity hotspots. Section Nintooa, widely distributed across mainland China, Vietnam, and Hainan Island, exhibits a typical disjunct distribution pattern. However, the origins of the flora of Hainan Island and the phylogenetic relationships within Section Nintooa remain contentious. In this study, we collected 81 samples encompassing all recognized species of Section Nintooa, along with samples from other sections of Lonicera. We reconstructed the phylogenetic relationships using 491 orthologous nuclear genes and complete plastomes generated by target enrichment and genome skimming sequencing. Our nuclear gene-based species and concatenated trees support the monophyly of Section Nintooa. The species tree indicates that the Vietnamese and Hainan populations form sister clades. However, the plastome results indicate that Section Nintooa is polyphyletic, with subsection Volubilis forming a monophyletic group and subsection Calcaratae forming a sister clade with other members of subgenus Chamaecerasus. Our hybridization network analysis reveals extensive gene flow within Section Nintooa, whereas subsection Calcaratae shows no gene flow with subsection Volubilis, leaving the origin of Calcaratae unclear. Lonicera species from Hainan Island exhibit unstable phylogenetic positions and underwent a rapid radiation during the Miocene. Biogeographical results indicate that populations from Vietnam and Hainan both originated from mainland China. Overall, our findings enhance the understanding of the evolutionary diversification history of Lonicera.

plant biology↗

bioRxiv2024

Olfactory cortical outputs recruit and shape distinct brain-wide spatiotemporal networks

Odor information is transmitted from the olfactory bulb to several primary olfactory cortical regions in parallel, including the anterior olfactory nucleus (AON) and piriform cortex (Pir). However, the specific roles of the olfactory bulb and cortical outputs in wider interactions with other interconnected regions throughout the brain remain unclear due to the lack of suitable in vivo techniques. Furthermore, emerging associations between olfactory-related dysfunctions and neurological disorders underscore the need for examining olfactory networks at the systems level. Using optogenetics, fMRI, and computational modeling, we interrogated the spatiotemporal properties of brain-wide neural interactions in olfactory networks. We observed distinct downstream recruitment patterns. Specifically, stimulation of excitatory projection neurons in OB predominantly activates primary olfactory network regions, while stimulation of OB afferents in AON and Pir primarily orthodromically activates hippocampal/striatal and limbic networks, respectively. Temporally, repeated OB or AON stimulation diminishes neural activity propagation brain-wide in contrast to Pir stimulation. Dynamic causal modeling analysis reveals a robust inhibitory effect of AON outputs on striatal and limbic network regions. In addition, experiments in aged rat models show decreased brain-wide activation following OB stimulation, particularly in the primary olfactory and limbic networks. Modeling analysis identifies a dysfunctional AON to Pir connection, indicating the impairment of this primary olfactory cortical circuit that disrupts the downstream long-range propagation. Our study delineates the spatiotemporal properties of olfactory neural activity propagation in brain-wide networks for the first time and distinguishes the roles of primary olfactory cortical, AON and Pir, outputs in shaping neural interactions at the systems level.

neuroscience↗

bioRxiv2024

A Multitrait Genome-Wide Association Study Reveals a Requirement for the Strigolactone Receptor DWARF14 in Optimal GOLVEN10 Signaling

GOLVEN/ROOT MERISTEM GROWTH FACTOR family of signaling peptides have been shown to control root lateral organ number, density and positioning in plants, although the signaling pathways involved remain obscure. A diverse set of 171 Medicago truncatula HapMap accessions with variation in responses to the GOLVEN 10 peptide, GLV10, were used to identify 74 significant loci controlling seven traits related to nodule formation and root architecture. Importantly, a single nucleotide polymorphism (SNP) in the upstream region of the MtGLV10 peptide-inducible strigolactone receptor gene, MtDWARF14 was significantly associated with insensitivity of nodule density to GLV10, suggesting a link between strigolactone signaling and GLV10 responsiveness. Three independent d14 mutants of the DWARF14 gene were found to hypernodulate, while overexpression of the gene led to reduction in nodule number, phenocopying GLV10. A null mutant, mtd14-1, remained sensitive to GLV10s effect on nodule density. However, at the transcriptional level, the mutant failed to effectively induce the expression of the GOLVEN marker genes, MtPLETHORA3 and MtPINLIKES2. Our study uncovers a hitherto unknown link between the strigolactone and GLV peptide signaling pathways using genotype x environment analysis of Medicago HapMap lines and provides a putative molecular mechanism for recovery from frost damage to fine roots.

plant biology↗

bioRxiv2024

Inverse Reinforcement Learning to Study Motivation in Mouse Behavioral Paradigms

Motivation describes the underlying goals that drive animal and agent behavior. In Neuroscience, behavioral paradigms are used to quantify the motivations of mice and used to gain insights into traits and diseases which can be translated to humans. In recent years, Computer Vision models are becoming widely adopted by Neuroscientists to score mouse behavior associated with motivations such as hunger and anxiety. However, a single motivation can be expressed by multiple different behaviors, and a single behavior can be linked to multiple motivations. Therefore the ideal analysis of motivational paradigms would attempt to directly recover the underlying motivations guiding behavior, rather than indirectly score their associated behaviors. In this paper, we move towards this goal by applying Inverse Reinforcement Learning to study the underlying motivations that drive mouse behavior.

animal behavior and cognition↗

bioRxiv2024

Cell-type specific inference from bulk RNA-sequencing data by integrating single cell reference profiles via EPIC-unmix

Cell type specific (CTS) analysis is essential to reveal biological insights obscured in bulk tissue data. However, single-cell (sc) or single-nuclei (sn) resolution data are still cost-prohibitive for large-scale samples. Thus, computational methods to perform deconvolution from bulk tissue data are highly valuable. We here present EPIC-unmix, a novel two-step empirical Bayesian method integrating reference sc/sn RNA-seq data and bulk RNA-seq data from target samples to enhance the accuracy of CTS inference. We demonstrate through comprehensive simulations across three tissues that EPIC-unmix achieved 4.6% - 109.8% higher accuracy compared to alternative methods. By applying EPIC-unmix to human bulk brain RNA-seq data from the ROSMAP and MSBB cohorts, we identified multiple genes differentially expressed between Alzheimers disease (AD) cases versus controls in a CTS manner, including 57.4% novel genes not identified using similar sample size sc/snRNA-seq data, indicating the power of our in-silico approach. Among the 6-69% overlapping, 83%-100% are in consistent direction with those from sc/snRNA-seq data, supporting the reliability of our findings. EPIC-unmix inferred CTS expression profiles similarly empowers CTS eQTL analysis. Among the novel eQTLs, we highlight a microglia eQTL for AD risk gene AP3B2, obscured in bulk and missed by sc/snRNA-seq based eQTL analysis. The variant resides in a microglia-specific cCRE, forming chromatin loop with AP3B2 promoter region in microglia. Taken together, we believe EPIC-unmix will be a valuable tool to enable more powerful CTS analysis.

genomics↗