Biology subjects

Wu, H.-J.

Publications and source records attributed to Wu, H.-J..

At least 19 recordsLinked to original sources

bioRxiv2026

Protein-based, live-cell PAINT microscopy with fluorogenic exchangeable HaloTag ligands

We report a generalizable method for single-molecule localization super-resolution microscopy in living cells. Live-cell PAINT (points accumulation for imaging in nanoscale topography) and single-molecule diffusivity mapping (SMdM) of intracellular protein targets are achieved by pairing exceptionally fluorogenic bis-trifluoromethyl rhodamine (BF) dyes with reversible HaloTag ligands and widely available HaloTag fusion proteins. This far-red small-molecule update to protein-based PAINT is readily incorporated into existing super-resolution microscopy workflows: wash-free staining and imaging are achieved for cell lines and primary neuron cultures for diverse protein targets. Pairing with photoconvertible fluorescent proteins further enables simultaneous two-color live-cell super-resolution microscopy and SMdM.

cell biology↗

bioRxiv2026

Evaluating Limits of Machine Learning-Assisted Raman Spectroscopy in Classification of Biological Samples

Machine learning (ML)-assisted Raman spectroscopy has become a powerful analytical tool for the classification and identification of analytes; however, technical challenges impacting its detection accuracy have not been investigated. This study explores experimental factors affecting classification performance. Among the evaluated ML models, ML algorithms show minimal impacts on classification accuracy. Instead, experimental factors, including spectral similarity between tested samples and the data quality, dominate detection performance. Increases in spectral noises and spectral similarity significantly reduce classification accuracy. In well-controlled samples with low experimental noise, ML-assisted Raman spectroscopy can discriminate lipid mixtures with a composition difference of 1.85 mol%. To assess the effect of biological heterogeneity, we analyzed single-cell Raman spectra from Saccharomyces cerevisiae strains carrying single, double, or triple gene mutations. Intrinsic cell-to-cell variability introduced substantial spectral differences, severely reducing the accuracy of multiclass classification of these genetically similar strains at the single-cell level. Averaging Raman spectra across multiple cells improved classification accuracy by reducing this spectral variability. We also assess the effectiveness of transfer learning across different Raman spectrometers, specifically by applying a ML model trained on one instrument to another Raman spectrometer. Transfer learning can be improved with proper instrument calibration, highlighting the importance of instrument standardization. Overall, our results demonstrate that data quality and spectral similarity are the primary bottlenecks in ML-assisted Raman spectroscopy. Careful attention to sample preparation, data acquisition, measurement conditions, and instrument calibration is critical to achieving robust and reliable classification performance.

bioinformatics↗

bioRxiv2025

Protein large language model assisted one-to-one gene homology mapping in cross-species single-cell transcriptome integration

Cross-species integration of single-cell transcriptomes requires establishing gene correspondences to enable comparative analysis of expression profiles across organisms. Current approaches predominantly rely on Ensembl homology tables; although gene-family expansion and contraction can reflect biologically meaningful evolutionary divergence, default many-to-many mappings can overweight expanded gene-family signals during integration and generate mapping-associated micro-clusters that lack clear cell-type identity, thereby complicating direct cell-type alignment. While restricting mappings to a one-to-one scheme suppresses such artifacts, it reduces the number of homology gene pairs by approximately 8% ([~]900 pairs). To address this limitation, we develop a protein large language model (pLLM)-based gene homology mapping strategy that boosts the number of homology gene pairs. By integrating pLLM-derived representations with sequence similarity, we construct a fused mapping approach, which achieves top performance in a comprehensive benchmark based on a curated cross-species atlas--spanning nine datasets, 11 species, and over 3.2 million cells. Our method further identifies previously unannotated cell-type marker pairs, facilitating novel cross-species marker discovery. These results establish a robust framework for gene homology mapping in cross-species transcriptome integration, improving both accuracy and biological interpretability.

bioinformatics↗

bioRxiv2025

Machine-learning strategy for high error tolerance in image-based digital molecular assays

There is a significant global health need to translate more in vitro diagnostic tests (IVDs) from clinical laboratories to field-based applications, including point-of-care (POC) and self-administered test formats. These applications typically require smaller sample sizes, limit the extent of sample processing and measurement capabilities, and introduce greater handling variability. Error tolerance is one of the most critical factors for successful field-based assay design. Here, we examine machine-learning (ML) strategies to enhance the error tolerance of image-based nanoparticle immunoassays. Random dispersions of nanoparticles were imaged in microliter sample volumes, and images were processed to determine analyte concentrations based on nanoparticle appearance. Assay performance was characterized using two common blood diagnostics: C-reactive protein (CRP) and S.CoV-2 IgG. We compare the results from a conventional image analysis, a hybrid ML-conventional approach based on pixel segmentation, and a full end-to-end image regression using a targeted regularization strategy. Training images for the full image regression approach required only a single label for training - the analyte concentration - eliminating the need for labor-intensive pixel-level labeling. Ultimately, the fully ML-based analysis significantly improved dynamic range, sensitivity, and reproducibility in high-error settings, including direct measurements performed in whole blood.

bioengineering↗

bioRxiv2025

Temporal control of sgRNA library activation unlocks large-scale in vivo CRISPR screens

Functional genomics screens have illuminated genetic dependencies in cancer, but conventional in vitro approaches fail to capture vulnerabilities shaped by the tumor microenvironment. Here, we implement CRISPR-StAR (Stochastic Activation by Recombination), a next-generation inducible CRISPR screening platform for large-scale in vivo applications. The system uses a dual lox-based recombination system to enable guide-level normalization and clonal knockout phenotyping. To analyze the rich (barcode-embedded) sequencing output, we developed UMIBB, a superior Bayesian statistical framework for quantifying gene-level dropout and enrichment compared to conventional software packages. Screening a 30,000-sgRNA library in A549 xenografts, followed by clone representation and dropout correlation analyses, showed high fidelity and reproducibility with dropout phenotypes resolvable using as few as 30 tumors for this size library. Validation across multiple tumor models demonstrated that a single tumor can provide reliable, functional annotation for [~]1,000 genes leveraging intra-tumor library controls for normalization. Comparing in vivo and in vitro screens revealed that a substantial subset of tumor suppressor genes exerts strong phenotypic effects only observable in vivo. For example, single-gene knockout and transcriptomic profiling confirmed that KMT2C and KMT2D have contrasting impacts on tumor growth - an insight that would have been overlooked in standard cell culture. Looking ahead, CRISPR-StAR screening, combined with our user-friendly analysis pipeline available on GitHub (R-package), offer an integrated framework for creating in vivo dependency maps that can complement existing vitro datasets like DepMap and Achilles. Critically, our approach reduces animal use by up to 7-fold compared to conventional in vivo dropout screens. This represents a significant ethical and methodological advancement - achieving genome-scale resolution with far fewer animals and greater reproducibility.

genomics↗

bioRxiv2025

Comprehensively Testing the Function of Missense Variation in the STK11 Tumour Suppressor

The tumor suppressor gene STK11 encoding Serine/Threonine Kinase 11 (STK11) is associated with Peutz-Jeghers Syndrome (PJS), a heritable gastrointestinal disease that increases lifetime cancer risk, and with somatic variation that contributes to [~]30% of lung and 20% of cervical cancers. Although identifying pathogenic variants is clinically actionable, over 94% of STK11 missense variants that have been observed clinically lack a definitive classification. We therefore measured the impact of STK11 variants at scale in a mammalian cell-based assay, scoring 6,026 (73% of all possible) amino acid substitutions across the full-length gene. Functional scores--which were consistent with biochemical properties, smaller-scale assays, and pathogenicity annotations--identified a subset of PJS patients with germline STK11 variants diagnosed later in life, as well as somatic STK11 variants found in cancer patients that had comparable overall survival estimates to wild-type STK11. Our scores provided new evidence for 350 annotated VUS STK11 missense variants and [~]80% of missense variants that have not yet been reported clinically, but we might expect to observe in the future. Thus, our effect map provides a proactive resource for gaining sequence-structure-function insights and evidence for actionable interpretation of clinical missense variants.

cancer biology↗

bioRxiv2025

Hi-Compass resolves cell-type chromatin interactions by single-cell and spatial ATAC-seq data across biological scales

Computational prediction of three-dimensional (3D) genome organization provides an alternative approach to overcome the cost and technical limitations of Hi-C experiments. However, current Hi-C prediction models are constrained by their narrow applicability to studying the impact of genetic variation on genome folding in specific cell lines, significantly restricting their biological utility. We present Hi-Compass, a generalizable deep learning model that accurately predicts chromatin organization across diverse biological contexts, from bulk to single-cell samples. Hi-Compass outperforms existing methods in prediction accuracy and is generalizable to unseen cell types through chromatin accessibility data, enabling broad applications in single cell omics. Hi-Compass successfully resolves cell-type-specific 3D genome architectures in complex biological scenarios, including immune cell states, organ heterogeneity, and tissue spatial organization. Furthermore, Hi-Compass enables integrative analysis of single-cell multiome data, linking chromatin interaction dynamics to gene expression changes across cell clusters, and mapping disease variants to pathogenic genes. Hi-Compass also extends to spatial multi-omics data, generating spatially resolved Hi-C maps that reveal domain-specific chromatin interactions linked to spatial gene expression patterns.

bioinformatics↗

bioRxiv2024

CellPatch: a Highly Efficient Foundation Model for Single-Cell Transcriptomics with Heuristic Patching

The rapid advancement of foundation models has significantly enhanced the analysis of single-cell omics data, enabling researchers to gain deeper insights into the complex interactions between cells and genes across diverse tissues. However, existing foundation models often exhibit excessive complexity, hindering their practical utility for downstream tasks. Here, we present CellPatch, a lightweight foundation model that leverages the strengths of the cross-attention mechanism and patch tokenization to reduce model complexity while extracting efficient biological representations. Comprehensive evaluations conducted on single-cell RNA-sequencing datasets across multiple organs and tissue states demonstrate that CellPatch achieves state-of-the-art performance in downstream analytical tasks while maintaining ultra-low computational costs during both pretraining and finetuning phases. Moreover, the flexibility and scalability of CellPatch allow it to serve as a general framework that can be incorporated with other well established single-cell analysis software, thereby enhancing their performance through transfer learning on diverse downstream tasks.

bioinformatics↗

bioRxiv2024

Spatiotemporal transcriptomic analyses reveal molecular gradient patterning during development and the tonotopic organization along cochlear axis

While cochlear tonotopic organization is essential for sound frequency discrimination, the spatiotemporal molecular programs shaping this architecture remain unresolved. Here, we integrate spatiotemporal transcriptomics (E17.5-P8-adult) with single-cell RNA-seq to construct a single-cell-resolution atlas of mouse cochlear development. As marked by differential gradient expression genes, distinct spatially and molecularly defined hair cell (HC) subtypes along the apical-basal axis were identified: such as the apical Myo7a+++/Calb1+++ outer hair cell (OHC) subtype, the basal Myo7a+/Calb1{square} OHC subtype, the apical vGlut3+++/Calb2+ inner hair cell (IHC) subtype, and the basal vGlut3+/Calb2+++ IHC subtype. In addition, the spatial data also demonstrate the developmental reversal of Myo7a/Calb2Nefh gradient patterns in HCs and spiral ganglion neurons (SGNs), revealing dynamic plasticity and complexity in frequency coding during development. Spatial analyses further demonstrate regional heterogeneity in cell communication intensity between HCs and SGNs, and also reveal core signaling hubs coordinating tonotopic organization, such as Fgf10-Fgfr2 pathway in the basal region and Ptn-Sdc pathway in the apical region HC or SGNs at P8 stage. Our study provides an open-access spatial database and reveals the morphological and molecular foundations underlying cochlear tonotopic organization, linking molecular and developmental mechanisms to auditory pathophysiology.

developmental biology↗

bioRxiv2024

Factors Promoting Lipopolysaccharide Uptake by Synthetic Lipid Droplets

Lipoproteins are essential in removing lipopolysaccharide (LPS) from blood during bacterial inflammation. The physicochemical properties of lipoproteins and environmental factors can impact LPS uptake. In this work, synthetic lipid droplets containing triglycerides, cholesterols, and phospholipids, were prepared to mimic lipoproteins. The physicochemical properties of these lipid droplets, such as charges, sizes, and lipid compositions, were altered to understand the underlying factors affecting LPS uptake. The amphiphilic LPS could spontaneously adsorb on the surface of lipid droplets without lipopolysaccharide binding protein (LBP); however, the presence of LBP can increase LPS uptake. The positively charged lipid droplets also enhance the uptake of negatively charged LPS. Most interestingly, the LPS uptake highly depends on the concentrations of Ca2+ near the physiological conditions, but the impact of Mg2+ ions was not significant. The increase of Ca2+ ions can improve LPS uptake by lipid droplets; this result suggested that Ca2+ may play an essential role in LPS clearance. Since septic shock patients typically suffer from hypocalcemia and low levels of lipoproteins, the supplementation of Ca2+ ions along with synthetic lipoproteins may be a potential treatment for severe sepsis.

biochemistry↗

bioRxiv2024

CRISPR screens reveal convergent targeting strategies against evolutionarily distinct chemoresistance in cancer

Resistance to chemotherapy has been a major hurdle that limits therapeutic benefits for many types of cancer. Here we systematically identify genetic drivers underlying chemoresistance by performing 30 genome-scale CRISPR knockout screens for seven chemotherapeutic agents in multiple cancer cells. Chemoresistance genes vary between conditions primarily due to distinct genetic background and mechanism of action of drugs, manifesting heterogeneous and multiplexed routes towards chemoresistance. By focusing on oxaliplatin and irinotecan resistance in colorectal cancer, we unravel that evolutionarily distinct chemoresistance can share consensus vulnerabilities identified by 26 second-round CRISPR screens with druggable gene library. We further pinpoint PLK4 as a therapeutic target to overcome oxaliplatin resistance in various models via genetic ablation or pharmacological inhibition, highlighting a single-agent strategy to antagonize evolutionarily distinct chemoresistance. Our study not only provides resources and insights into the molecular basis of chemoresistance, but also proposes potential biomarkers and therapeutic strategies against such resistance.

cancer biology↗

bioRxiv2024

STCC: consensus clustering enhances spatial domain detection for spatial transcriptomics data

The rapid advance of spatially resolved transcriptomics technologies has yielded substantial spatial transcriptomics data. Deriving biological insights from these data poses non-trivial computational and analysis challenges, of which the most fundamental step is spatial domain detection (or spatial clustering). Although a number of tools for spatial domain detection have been proposed in recent years, their performance varies across datasets and experimental platforms. It is thus an important task to take full advantage of different tools to get a more accurate and stable result through consensus strategy. In this work, we developed STCC, a novel consensus clustering framework for spatial transcriptomics data that aggregates outcomes from state-of-the-art tools using a variety of consensus strategies, including Onehot-based, Average-based, Hypergraph-based and wNMF-based methods. Comprehensive assessments on simulated and real data from distinct experimental platforms show that consensus clustering significantly improves clustering accuracy over individual methods under varied input parameters. For normal tissue samples exhibiting clear layered structure, consensus clustering by integrating multiple baseline methods leads to improved results. Conversely, when analyzing tumor samples that display scattered cell type distribution patterns, integration of a single baseline method yields satisfactory performance. For consensus strategies, Average-based and Hypergraph-based approaches demonstrated optimal precision and stability. Overall, STCC provides a scalable and practical solution for spatial domain detection in spatial transcriptomic data, laying a solid foundation for future research and applications in spatial transcriptomics.

bioinformatics↗

bioRxiv2023

A deep learning-based drug repurposing screening and validation for anti-SARS-CoV-2 compounds by targeting the cell entry mechanism

The recent outbreak of Corona Virus Disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been a severe threat to the global public health and economy, however, effective drugs to treat COVID-19 are still lacking. Here, we employ a deep learning-based drug repositioning strategy to systematically screen potential anti-SARS-CoV-2 drug candidates that target the cell entry mechanism of SARS-CoV-2 virus from 2,635 FDA-approved drugs and 1,062 active ingredients from Traditional Chinese Medicine herbs. In silico molecular docking analysis validates the interactions between the top compounds and host receptors or viral spike proteins. Using a SARS-CoV-2 pseudovirus system, we further identify several drug candidates including Fostamatinib, Linagliptin, Lysergol and Sophoridine that can effectively block the cell entry of SARS-CoV-2 variants into human lung cells even at a nanomolar scale. These efforts not only illuminate the feasibility of applying deep learning-based drug repositioning for antiviral agents by targeting a specified mechanism, but also provide a valuable resource of promising drug candidates or lead compounds to treat COVID-19.

microbiology↗

bioRxiv2023

Spider: a flexible and unified framework for simulating spatial transcriptomics data

Spatial transcriptomics technology provides a valuable view for studying cellular heterogeneity due to its ability to simultaneously acquire gene expression profile and cell location information. However, benchmarking these rapidly accumulating spatial transcriptomics analysis tools is challenging owing to the limited diversity and accuracy of "gold standard" data sets annotated by pathologists. To address this issue, we proposed Spider, a flexible and unified simulator for spatial transcriptomics data guided by cell type proportion and transition matrix of adjacent cell types. Taking advantage of a heuristic batched simulated annealing algorithm (BSA) in assigning simulated cell type labels, Spider can generate spatial transcriptomics data for one million cells in just five minutes. Furthermore, Spider can generate various types of spatial transcriptomics data, including immune hot/cold tumor samples by specifying different immune cell proportions and transition matrices and layered tissue samples via an interactive interface. In addition, Spider is also a unified framework for ST data simulation in which we have implemented diverse simulators proposed by other researchers as special cases. We have systematically evaluated the performance of Spider and competing tools, and demonstrated Spiders remarkable power to capture the spatial pattern of the reference dataset. Spider is available at https://github.com/YANG-ERA/Artist.

bioinformatics↗

bioRxiv2023

Advancing Glycan Analysis: A New Platform Integrating SERS, Boronic Acids, and Machine Learning Algorithms

Glycans are the most abundant and diverse fundamental biomolecules. Profiling glycans is essential to elucidate biomolecular mechanisms and to predict biological consequences; however, the analysis of glycans remains challenging due to their structural complexity. A novel glycan detection platform is established through the integration of surface-enhanced Raman spectroscopy (SERS), boronic acid receptors, and machine learning tools. SERS is able to distinguish isomeric structures and provide fingerprint information of molecules. Boronic acid receptors can selectively bind with glycans, and the reaction influences molecular vibrations, leading to unique Raman spectral patterns. For proof-of-concept, two boronic acids, 4-mercaptophenylboronic acid (4MBA) and 1-thianthrenylboronic acid (1TBA) were used for glycan detection. The spectra were analyzed by the machine learning algorithm. The sensing platform successfully recognized the stereoisomers (glucose, mannose, and galactose), the structural isomers with -1,4 and -1,6 glycosidic linkages, and the {beta}-1,4 glycosidic linkage within lactose molecules. The collective spectra that combine the spectra from both boronic acid receptors could improve the performance of the support vector machine model due to the enrichment of the structural information of glycans. In addition to qualitative analysis, this new sensor could quantify the mole fraction of sialic acid in lactose background using the machine learning regression technique. The quantification accuracy reached a coefficient of determination (R2) value of 0.998 and a normalized mean square error (NMSE) of 0.00195. This low-cost, rapid, and highly accessible sensor will provide the scientific community with another option for frequent glycan screening in standard biological laboratories.

bioengineering↗

bioRxiv2022

Discovery and characterization of LNCSOX17 as an essential regulator in human endoderm formation

Long non-coding RNAs (lncRNAs) have emerged as fundamental regulators in various biological processes, including embryonic development and cellular differentiation. Despite much progress over the past decade, the genome-wide annotation of lncRNAs remains incomplete and many known non-coding loci are still poorly characterized. Here, we report the discovery of a previously not annotated lncRNA that is transcribed upstream of the SOX17 gene and located within the same topologically associating domain. We termed it LNCSOX17 and show that it is induced following SOX17 activation but its expression is more tightly restricted to early definitive endoderm. Loss of LNCSOX17 affects crucial functions independent of SOX17 and leads to an aberrant endodermal transcriptome, signaling pathway deregulation and epithelial to mesenchymal transition defects. Consequently, cells lacking the lncRNA cannot further differentiate into more mature endodermal cell types. We identified and characterized LNCSOX17 as an essential new actor in early human endoderm, thereby further expanding the list of functionally important non-coding regulators.

genomics↗

bioRxiv2022

Resolving clonal substructure from single cell genomic data using CopyKit

High-throughput methods for single cell copy number sequencing have enabled the profiling of thousands of cells in parallel, yet there remains a significant bottleneck for data analysis. Here we present CopyKit, a comprehensive set of computational methods for the pre-processing and analysis of single cell copy number data to resolve clonal substructure and reconstruct genetic lineages in tumors. We performed single cell DNA sequencing of 2977 cells from multiple spatial regions in two liver metastasis and 7365 cells from three primary tumors with matched metastatic tissues. In the liver metastases, CopyKit resolved clonal substructure in different spatial regions, which revealed both clonal intermixing and spatial segregation in the tumor mass. In the matched metastatic colorectal and breast cancers, CopyKit resolved metastatic lineages and identified subclones and genomic events that were associated with metastases. These applications show that CopyKit is comprehensive tool for resolving copy number substructure in tumors.

genomics↗

bioRxiv2022

Assessing base-resolution DNA mechanics on the genome scale

Intrinsic DNA properties such as bending play a crucial role in diverse biological systems. A recent advantage in the high-throughput method called loop-seq makes it possible to determine bendability of hundred thousand 50-bp DNA duplexes in one experiment. However, its still infeasible to assess whole sequence bendability in large genomes such as human, which needs thousands of loop-seq experiments. Here we introduce BendNet - a neural network to accurately predict the intrinsic DNA bending at base-resolution by only given DNA sequences. BendNet can increase the resolution of experimental results, and can predict DNA bendability for any new given sequences in high accuracy. We applied BendNet to the human genome and observed high-stiffness regions located at both transcriptional start sites and transcriptional end sites. Such stiffness patterns are different for coding and non-coding genes, which matches distinct nucleosome occupancy patterns. As expected, most transcription factors (TFs) bind in DNA of low bendability. In contrast, we observed an unusually high bendability within binding elements of specific TFs such as EBF1 and regulators of genome folding such as CTCF. These factors either co-bind or compete with nucleosomes to carry out their functions. More interestingly, CTCF binding regions exhibit the highest bendability than other DNA elements, implying their potential role in trapping and holding the CTCF in the exact locations to make sure CTCF as stable anchor in loop extrusion process. Our work provides a tool to assess DNA bendability for large-scale DNA sequences and expands our understanding on DNA mechanics in chromatin regulation and genome folding.

genomics↗