Biology subjects

Deng, P.

Publications and source records attributed to Deng, P..

At least 19 recordsLinked to original sources

bioRxiv2026

RegDesign-Bench: Benchmarking computational methods for cell-type-specific regulatory DNA design

The de novo design of regulatory DNA sequences that drive cell-type-specific gene expression could enable precise gene control in synthetic biology, cell engineering and gene therapy. Various computational methods have been developed for this task, yet systematic comparisons that guide method selection for different design objectives remain limited. Here, we introduce RegDesign-Bench, a standardized framework for evaluating cell-type-specific cis-regulatory element design through a comprehensive set of computational analyses. Among the methods tested, strong predicted activity was not consistently accompanied by cell-type-selective accessibility or prominent cell-type-associated motif signals. Similarity to reference sequence distributions also did not ensure high predicted target-cell activity or specificity. These findings demonstrate that favourable performance on individual measurements does not necessarily extend across regulatory properties and library characteristics, revealing application-dependent strengths and limitations of current design methods. Overall, RegDesign-Bench offers a general-purpose computational framework for comparing and selecting methods to design regulatory DNA that drives cell-type-specific gene expression.

genomics↗

bioRxiv2026

Reconstructing sequence-grammar trajectories enables interpretable and tunable cis- regulatory element design

Designing synthetic cis-regulatory elements (CREs) with cell-type-specific activity remains challenging, and optimization is usually treated as a black box, obscuring how regulatory grammar emerges and why design trajectories fail. Here, we present GO-CRE (Guided Optimization of Cis-Regulatory Elements), which combines efficient sequence generation, predictor-guided reinforcement learning, and trajectory-level interpretation. GO-CRE reconstructs iterative sequence changes in a shared sequence-grammar landscape and identifies coordinated update programs corresponding to search, commitment, and optimization. Productive trajectories in HepG2 and K562 progressively acquired cell-type-associated grammar, whereas SK-N-SH trajectories remained confined to local basins. In HepG2, trajectory analysis revealed a low-complexity polyG trap; introducing a polyG penalty redirected optimization toward HNF/FOXA-associated features. Final designs retained sequence diversity while converging on cell-type-associated motif patterns. Lentiviral MPRA validated cell-type-specific activity in K562 and HepG2 and showed higher average activity of HepG2 designs than endogenous CREs. Together, these findings establish sequence-grammar trajectory reconstruction as a basis for interpretable and tunable synthetic CRE design.

genomics↗

bioRxiv2026

HoloCell: A Generative Foundation Model for Holistic Cellular Modeling

Single-cell multi-omics technologies have recently advanced to enable the profiling of epigenomic, transcriptomic, and proteomic layers within individual cells, offering new opportunities to characterize cellular states as integrated biological systems. However, developing a unified framework that can seamlessly integrate diverse omics modalities and remain robust to heterogeneous modality missingness remains challenging. Existing methods are often designed for specific modalities or modality pairs, relying on dataset-specific training or paired measurements. Here we present HoloCell, to our knowledge the first generative foundation model for joint representation learning and generative modeling across all three major single-cell omics modalities, i.e., epigenomics, transcriptomics, and proteomics. HoloCell contains over 860 million parameters and is pretrained on the Human-Multi-Omics-Corpus, which comprises approximately 468 million single-cell profiles across these three omics layers, corresponding to over 425 billion tokens. HoloCell introduces a a simple yet biologically motivated hierarchical tokenization strategy that encodes cis-regulatory elements, genes, and proteins as structured tokens within a shared modeling framework. We evaluated HoloCell across single-omics representation learning, paired multi-omics integration, unpaired multi-omics alignment, and cross-modal generation via iterative diffusion and remasking, demonstrating its superior performance and flexibility across diverse omics tasks. From a representation perspective, HoloCell provides a unified digital mapping of cellular states across multiple omics layers, capturing cell heterogeneity as an integrated system. From a generation perspective, its iterative diffusion and remasking frame-work permits flexible generation orders beyond fixed left-to-right causality, enabling in silico simulation of multi-omics information flow. Together, these capabilities position HoloCell as a versatile foundation model toward the emerging concept of a virtual cell, offering both systematic characterization and generative simulation of cellular systems within a unified framework.

bioinformatics↗

bioRxiv2026

Phenotype-Guided In Silico Molecular Generation Using Large Language Models

Complex diseases often emerge from coordinated, system-level cellular state changes that are difficult to address with target centric drug discovery. Phenotypic drug discovery offers a principled alternative but remains constrained by the cost and scalability of pathologically relevant assays. Here we present GEMGen, a large language model-based framework that performs in silico phenotypic drug discovery by generating small molecules directly from transcriptomic representations of cellular states. GEMGen encodes desired phenotypic transitions as text-based representations of up- and down-regulated gene sets, enabling transferable modeling across experimental platforms and data modalities. Trained on large scale chemical perturbation data, GEMGen robustly identifies phenotype-oriented compounds and mechanistically related but structurally distinct candidates across multiple benchmarks. Applied to signatures induced by genetic perturbations, GEMGen produces small molecules that phenocopy gene knockdown effects and identifies chemically novel inhibitors, including previously unreported KEAP1 inhibitors that activate NRF2 signaling. Extending this approach to a disease relevant model of fibrosis, GEMGen generates compounds that reverse profibrotic transcriptional programs and cellular phenotypes. These results establish a scalable framework for translating transcriptomic phenotypes into candidate therapeutic molecules, enabling systems-level exploration of vast chemical space and offering a complementary in silico counterpart to physical phenotypic drug screens.

cell biology↗

bioRxiv2025

Regularized Single-cell Imaging Enables Generalizable AI models for Stain-free Cell Viability Screening

Cell viability assays are essential tools in biomedical research and drug development. Artificial intelligence (AI) offers the potential to simplify these assays by predicting cell viability directly from brightfield microscopy images, but current models lack generalizability across diverse cell types and treatments. Here, we introduce a strategy called "regularized imaging", where single cells are isolated in nanowells to generate standardized image patches that simplify segmentation and improve training data quality. We trained our model using example images of live and dead cells from a single cell line exposed to four cytotoxic conditions (ethanol, andrographolide, daunorubicin, and serum starvation). Despite this narrow training dataset, the resulting model accurately identified live and dead cells after treatment by previously unseen compounds, successfully replicating dose-response curves comparable to fluorescence live/dead assays. Importantly, this model effectively generalized across diverse cell types, including both adherent and suspension cells. Additionally, microscopy-based cell viability analysis is non-destructive, enabling repeated measurements to perform kinetic studies to distinguish between fast- and slow-acting compounds. Our findings highlight how regularized single cell imaging enables the training of broadly generalizable AI models to recognize biologically relevant cell features for label-free cell screening workflows. One-sentence SummaryRegularized single-cell imaging in nanowells enables training of generalizable AI models for accurate, stain-free viability screening across previously unseen cell types and contexts.

bioengineering↗

bioRxiv2025

High-throughput Single Cell Motility Analysis using Nanowell-in-Microwells

Cell motility is important to many biological processes including cancer, immune response, and tissue repair. Conventional assays measure bulk cell motility, potentially overlooking important heterogeneity and missing important high motility subpopulations. Here, we introduce a high-throughput single-cell motility assay using nanowell-in-microwell plates to precisely track single cell position and analyze their migratory trajectories. By physically confining individual cells in nanowells, we eliminate cell-cell interactions and simplify cell segmentation and tracking. Using this platform, we characterized the motility of single cells across different culture conditions to identify distinct motility phenotypes. Single-cell trajectory analysis revealed pronounced directional persistence, with cells predominantly maintaining their direction of travel and trajectory along nanowell boundaries. Additionally, our approach facilitates the generation of labeled image datasets suitable for AI models to rapidly identify cell motility phenotypes from single-cell images. Together, our platform provides a robust, scalable method to analyze cell motility phenotypes and migration behavior at single-cell resolution. Significance StatementA novel single-cell motility assay that uses confinement in nanowell-in-microwell to enable high-throughput profiling of single cell motility phenotypes.

bioengineering↗

bioRxiv2025

Ketogenic diet dampens excitatory neurotransmission by shrinking synaptic vesicle pools

HIGHLIGHTSO_LIKD rewires gene expression in the hippocampus, particularly impacting synaptic genes. C_LIO_LIKD profoundly alters epigenetic modifications in the hippocampus. C_LIO_LIKD reduces excitatory synaptic gain and dampens integration of synaptic inputs. C_LIO_LIKD reduces the readily releasable vesicle pool in excitatory synapses. C_LI The ketogenic diet (KD) is a common dietary intervention for treating seizures in intractable childhood epilepsies and has been proposed to improve disease outcome in neurodegenerative disorders. Despite its clinical applications, we know little about how this diet impacts brain circuitry and neuronal function to elicit its protective effects. Here, we examined the impact of the KD on hippocampal function through integrative analysis of gene expression, epigenetics and neurotransmission. We found that KD induces profound transcriptional reprogramming of the hippocampus, including dampened expression of numerous synaptic genes. Through proteomic analysis of histone variants and post-translational modifications, we uncovered significant changes in activating and repressive histone marks in the hippocampus of KD mice. To determine how transcriptional rewiring of the hippocampus under KD impacts neurotransmission, we performed electrophysiological recordings of neurotransmission and synaptic dynamics at excitatory CA3-CA1 synapses. We found that KD diminishes synaptic gain and dampens short-term plasticity at excitatory synapses, resulting in reduced integration of synaptic inputs at the circuit level. Combining electrophysiology and electron microscopy, we determined that effects of KD in excitatory synapses are caused by a reduction in size of the readily releasable pool of synaptic vesicles, as well as the total vesicle pool. Our findings show that the ketogenic diet triggers synaptic remodeling in the hippocampus, driven by broad transcriptional and epigenetic changes that reduce synaptic vesicle pools and short-term plasticity at excitatory synapses ultimately dampening excitatory synaptic gain and integration at the circuit level. These synaptic adaptations may represent a major mechanism underlying the anti-epileptic effects of this diet.

neuroscience↗

bioRxiv2025

Nurr1 orchestrates claustrum development and functionality

Claustrum is the part of the forebrain known to be most widely interconnected tissue with almost all global regions. It is believed to be involved in coordinating multiple cognitive behaviors including consciousness formation. However, little is known about the molecular mechanisms underlying its development and involvement in behavioral control. Here we show that Nurr1 (Nr4a2) is the key transcription factor orchestrating claustral morphogenesis, functional connectivity and thus behaviors. Nurr1 deficient claustral cells aberrantly migrate into insular cortex, shaping claustrum into an abnormal wing-like structure, and ectopically turn on insular corticex genetic program. Accordingly, functional connectivity of the claustrum is not properly formed and relevant behaviors are dysregulated in Nurr1 deficient mice. We show that Nurr1 regulates claustral neuron positioning by suppressing G-protein signaling.

neuroscience↗

bioRxiv2024

Directing cellular transitions on gene graph-enhanced cell state manifold

A select few genes act as pivotal drivers in the process of cell state transitions. However, finding key genes involved in different transitions is challenging. To address this problem, we present CellNavi, a deep learning-based framework designed to predict genes that drive cell state transitions. CellNavi builds a driver gene predictor upon a cell state manifold, which captures the intrinsic features of cells by learning from large-scale, high-dimensional transcriptomics data and integrating gene graphs with directional connections. Our analysis shows that CellNavi can accurately predict driver genes for transitions induced by genetic, chemical, and cytokine perturbations across diverse cell types, conditions, and studies. By leveraging a biologically meaningful cell state manifold, it is proficient in tasks involving critical transitions such as cellular differentiation, disease progression, and drug response. CellNavi represents a substantial advancement in driver gene prediction and cell state manipulation, opening new avenues in disease biology and therapeutic discovery.

cell biology↗

bioRxiv2024

Structure and mechanism of biosynthesis of Streptococcus mutans cell wall polysaccharide

Streptococcus mutans, the causative agent of human dental caries, expresses a cell wall attached Serotype c-specific Carbohydrate (SCC) that is critical for cell viability. SCC consists of a repeating [->]3)-Rha(1[->]2)-Rha(1[->] polyrhamnose backbone, with glucose (Glc) side-chains and glycerol phosphate (GroP) decorations. This study reveals that SCC has one major and two minor Glc modifications. The major Glc modification, -Glc, attached to position 2 of 3-rhamnose, is installed by SccN and SccM glycosyltransferases and is the site of the GroP addition. The minor Glc modifications are {beta}-Glc linked to position 4 of 3-rhamnose installed by SccP and SccQ glycosyltransferases, and -Glc attached to position 4 of 2-rhamnose installed by SccN working in tandem with an unknown enzyme. Both the major and the minor {beta}-Glc modifications control bacterial morphology, but only the GroP and major Glc modifications are critical for biofilm formation.

microbiology↗

bioRxiv2024

Target-aware Molecule Generation for Drug Design Using a Chemical Language Model

Generative drug design facilitates the creation of compounds effective against pathogenic target proteins. This opens up the potential to discover novel compounds within the vast chemical space and fosters the development of innovative therapeutic strategies. However, the practicality of generated molecules is often limited, as many designs focus on a narrow set of drug-related properties, failing to improve the success rate of subsequent drug discovery process. To overcome these challenges, we develop TamGen, a method that employs a GPT-like chemical language model and enables target-aware molecule generation and compound refinement. We demonstrate that the compounds generated by TamGen have improved molecular quality and viability. Additionally, we have integrated TamGen into a drug discovery pipeline and identified 7 compounds showing compelling inhibitory activity against the Tuberculosis ClpP protease, with the most effective compound exhibiting a half maximal inhibitory concentration (IC50) of 1.9 M. Our findings underscore the practical potential and real-world applicability of generative drug design approaches, paving the way for future advancements in the field.

biochemistry↗

bioRxiv2023

Accelerating protein engineering with fitness landscape modeling and reinforcement learning

Protein engineering holds significant promise for designing proteins with customized functions, yet the vast landscape of potential mutations versus limited lab capacity constrains the discovery of optimal sequences. To address this, we present the {micro}Protein framework, which accelerates protein engineering by combining {micro}Former, a deep learning model for accurate mutational effect prediction, with {micro}Search, a reinforcement learning algorithm designed to efficiently navigate the protein fitness landscape using {micro}Former as an oracle. {micro}Protein leverages single mutation data to predict optimal sequences with complex, multi-amino acid mutations through its modeling of epistatic interactions and a multi-step search strategy. Except from state-of-the-art performance on benchmark datasets, {micro}Protein identified high-gain-of-function multi-point mutants for the enzyme {beta}-lactamase, surpassing the highest known activity level, in wet-lab, trained solely on single mutation data. These results demonstrate {micro}Proteins capability to discover impactful mutations across vast protein sequence space, offering a robust, efficient approach for protein optimization.

synthetic biology↗

bioRxiv2023

Circuit-based intervention corrects excessive dentate gyrus output in the Fragile X mouse model

Abnormal cellular and circuit excitability is believed to drive many core phenotypes in fragile X syndrome (FXS). The dentate gyrus is a brain area performing critical computations essential for learning and memory. However, little is known about dentate circuit defects and their mechanisms in FXS. Understanding dentate circuit dysfunction in FXS has been complicated by the presence of two types of excitatory neurons, the granule cells and mossy cells. Here we report that loss of FMRP markedly decreased excitability of dentate mossy cells, a change opposite to all other known excitability defects in excitatory neurons in FXS. This mossy cell hypo-excitability is caused by increased Kv7 function in Fmr1 KO mice. By reducing the excitatory drive onto local hilar interneurons, hypo-excitability of mossy cells results in increased excitation/inhibition ratio in granule cells and thus paradoxically leads to excessive dentate output. Circuit-wide inhibition of Kv7 channels in Fmr1 KO mice increases inhibitory drive onto granule cells and normalizes the dentate output in response to physiologically relevant theta-gamma coupling stimulation. Our study suggests that circuit-based interventions may provide a promising strategy in this disorder to bypass irreconcilable excitability defects in different cell types and restore their pathophysiological consequences at the circuit level.

neuroscience↗

bioRxiv2023

Genetic basis of expression and splicing underlying spike architecture in wheat (Triticum aestivum L.)

IntroductionWheat is one of the most important staple crops worldwide, and an important source of human protein and mineral element intake. Continuously increasing stable production of wheat is critical for global food security under the challenge of population growth and limited resource input. ObjectiveSpike architecture determines the potential grain yield of wheat. However, the mechanisms of transcriptional regulation of spike architecture in wheat remain largely unknown, limiting further genetic improvement of wheat yield. In this study we explored the genetic basis of spike architecture in wheat. MethodsPopulation RNA-seq methods were used to identify the eQTLs and sQTLs associated with spike architecture and applied this to dissection of the genetic basis of gene expression and splicing controlling these complex yield-related traits. ResultsIn total, 4,143 expression quantitative trait loci (eQTLs) and 12,933 splice QTLs (sQTLs) were identified in wheat based on 178 RNA-seq samples, revealing 774 cis-eQTLs and 321 cis-sQTLs for 86 eGenes and 73 sGenes, respectively. Integration of eQTLs and sQTLs with genome-wide association study (GWAS) identified dozens of additional novel candidate genes that may contribute to spike-related traits. Gene network analysis showed that eQTLs and sQTLs were widely involved in the co-expression modules that regulate wheat spike architecture. Notably, the eQTL locus AX-108754757 regulated the expression of 5 eGenes that negatively controled grain number per spike. AX-111592099 regulated both the splicing and expression of TraesCS7B02G442100, encoding an E3 ubiquitin ligase, and playing a central role in regulating spike length. ConclusionThis study provides new insights into the genetic basis of spike architecture. This improved understanding of spike-related traits in wheat will contribute to more rapid genetic improvement.

plant biology↗

bioRxiv2023

Structural RNA components supervise the sequential DNA cleavage in R2 retrotransposon

Retroelements are the widespread jumping elements considered as major drivers for genome evolution, which can also be repurposed as gene-editing tools. Here, we determined the cryo-EM structures of eukaryotic R2 retrotransposon with ribosomal DNA target and regulatory RNAs. Combined with biochemical and sequencing analysis, we revealed two essential DNA regions, Drr and Dcr, required for R2 recognition and cleavage. The association of 3 regulatory RNA with R2 protein accelerates the first-strand cleavage, blocks the second-strand cleavage, and initiates the reverse transcription starting from the polyA tail. Removing 3 regulatory RNA by reverse transcription allows the association of 5 regulatory RNA and initiates the second-strand cleavage. Our work explained the DNA recognition and supervised sequential retrotransposition mechanisms by R2 machinery, providing novel insights into the retrotransposon and application reprogramming.

molecular biology↗

bioRxiv2023

Modeling cross-cell type cis-regulatory patterns via hierarchical deep neural network and gene expression prediction

Linking cis-regulatory sequences to target genes has been a long-standing challenge. In this study, we introduce CREaTor, an attention-based deep neural network designed to model cis-regulatory patterns for genomic elements up to 2Mb from target genes. Coupled with a training strategy that predicts gene expression from flanking candidate cis-regulatory elements (cCREs), CREaTor can model cell type-specific cis-regulatory patterns in new cell types without prior knowledge of cCRE-gene interactions or additional training. The zero-shot modeling capability, combined with the use of RNA-seq and ChIP-seq data only, allows for the readily generalization of CREaTor to a broad range of cell types. Evaluation reveals that CREaTor outperforms existing methods in capturing cCRE-gene interactions across various distance ranges in held-out cell types. Further analysis indicates that the superior performance of CREaTor can be attributed to its capacity to model regulatory interactions at multiple levels, including the higher-order genome organizations that govern cCRE activities as well as cCRE-gene interactions. Collectively, our findings highlight CREaTor as a powerful tool for systematically investigating cis-regulatory programs across various cell types, both in normal developmental processes and disease-associated contexts.

genetics↗

bioRxiv2022

NAD+ biosynthesis as a collateral lethality target for precision oncology

Genomic deletion of tumor suppressor genes (TSG) often encompasses neighboring genes which may be members of multi-gene families encoding cell essential functions. These genomic events create targetable cancer-specific vulnerabilities, termed "collateral lethality" as illustrated by homozygous deletion of the glycolytic gene ENO1, which sensitizes glioblastoma (GBM) cells to inhibition of its paralog ENO2. Here, we sought to generalize the concept by validating a second multi-gene family in an unrelated metabolic pathway. Nicotinamide-nucleotide adenylyltransferase (NMNAT) is an essential and (unlike the more extensively studied NAMPT) non-bypassable step in NAD biosynthesis encoded by three paralogs, one of which, NMNAT1, is homozygously deleted as part of the 1p36 tumor suppressor locus in TCGA data of GBM, Cholangiocarcinoma and Hepatocellular carcinoma (with near zero expression of found in several other malignancies). In a glioma cell line (Gli56) with homozygous deletion of NMNAT1 expressing NMNAT2 and NMNAT3, shRNA-mediated knockdown of NMNAT2 is selectively toxic to Gli56 NMNAT1-deleted but not ectopically rescued cells. As NMNAT1 and NMNAT2 are predominantly localized in the nucleus and cytosol, respectively, these data suggest a functionally common pool of cytosolic and nuclear NAD+. Inducible shRNA-mediated extinction of NMNAT2 decreases NAD+ levels and selectively kills NMNAT1-deleted, but not NMNAT1-rescued cells in vitro and eradicates intracranial tumors in vivo. Thus, collateral lethality is a generalizable framework for the development of new classes of targeted agents with an informed clinical development path in cancer. Statement of significanceAn ongoing challenge in precision oncology is the translation of genomic data into actionable therapeutic opportunities with clear clinical benefit. We have demonstrated that genes homozygously deleted by virtue of chromosomal proximity to major tumor suppressor genes can confer cancer-specific vulnerabilities, termed "collateral lethality". While our previous work validated one such collateral lethality target in glycolysis, we now provide empirical in vitro and in vivo evidence that this concept applies to another deleted gene governing an altogether distinct biochemical pathway. The generalization of collateral lethality may expand the spectrum of molecular targets in cancer with a genomically-informed path for accurate clinical development.

cancer biology↗

bioRxiv2021

Wall-associated Receptor Kinase and The Expression Profiles in Wheat Responding to Fungal Stress

Cell wall-associated kinases (WAKs), which are encoded by conserved gene families in plants, are crucial for development and responses to diverse stresses. However, the wheat (Triticum aestivum L.) WAKs have not been systematically classified, especially those involved in protecting plants from disease. Here, we classified 129 WAK proteins (encoded by 232 genes) and 75 WAK-Like proteins (WAKLs; encoded by 109 genes) into four groups, via a phylogenetic analysis. An examination of protein sequence alignment revealed diversity in the GUB-domain of WAKs structural organization, but it was usually characterized by a PYPFG motif followed by CxGxGCC motifs, while the EGF-domain was usually initiated with a YAC motif, and eight cysteine residues were spliced by GNPY motif. The expression profiles of WAK-encoding homologous genes varied in response to Blumeria graminis f. sp. tritici (Bgt), Puccinia striiformis f. sp. tritici (Pst) and Puccinia triticina (Pt) stress. A quantitative real-time polymerase chain reaction (qRT-PCR) analysis proved that TaWAK75 and TaWAK76b were involved in wheat resistance to Bgt. This study revealed the structure of the WAK-encoding genes in wheat, which may be useful for future functional elucidation of wheat WAKs responses to fungal infections.

plant biology↗