Yuehua Cui

Michigan State University
Country flagUnited States

Research Interests

ApplyKite TwinLearns as you apply

Make your next move.

Prepare your application

Review drafts before you send them.

Explore related searches

Articles (20)

With advances in spatially resolved transcriptomics across platforms and resolutions, it is now possible to measure gene-expression profiles while preserving the tissue microenvironment. Spatial clustering is central to these analyses, and recent graph neural network (GNN)-based approaches have greatly improved their accuracy. Nevertheless, precisely delineating spatial domain boundaries remains a significant challenge. Here we present a novel method CASTLE (Cell-type Aware SpaTial domain detection via contrastive Learning Embedding) by leveraging cell type information for improved spatial domain detection. CASTLE first integrates spatial proximity to construct a spatial graph, refined from cell-type similarity or expression similarity when cell type information is unavailable. Then, a self-supervised, local-context contrastive objective aligns embeddings with their immediate microenvironments. Evaluated across multiple tissues and technology resolutions, CASTLE outperforms state-of-the-art methods and detects more accurate and stable spatial domains. In downstream analyses, it resolves fine-grained tissue structures and differentiates functional subtypes, supporting more nuanced biological interpretation.

Year:

2026

Spatial transcriptomics (ST) measures gene expression while preserving spatial context within tissues. One of the key tasks in ST analysis is spatial domain detection, which remains challenging due to the complex structure of ST data and the varying performance of individual clustering methods. To address this, we propose SpatialESD, a Spatial EnSemble Domain detection method that integrates results from different spatial domain detection methods to improve spatial domain detection. SpatialESD captures both direct cooccurrence patterns and multiscale indirect relationships between clusters, improving the robustness and accuracy of spatial domain detection. We evaluated SpatialESD on simulated datasets and multiple 10x Visium spatial transcriptomics datasets, including human brain, breast cancer, and ovarian cancer samples. The results show that SpatialESD consistently outperforms individual methods and the existing EnSDD ensemble method in terms of clustering accuracy and stability. Based on the identified domains, we further detected domain‐specific differentially expressed genes and performed trajectory and cell–cell interaction analyses. These results reveal spatial patterns of gene expression and cellular communication, offering insights into tissue organization and disease mechanisms. Overall, SpatialESD provides a reliable and effective solution for spatial domain detection in ST data and facilitates downstream biological discovery.

Year:

2026

Spatial transcriptomics (ST) technologies offer unprecedented insights into the spatial organization of gene expression, allowing for the study of tissue architecture, domain boundaries, and cell–cell interactions. However, most ST data generated so far are at multicellular resolution, where each spot captures transcripts from a mixture of diverse cells of different cell types. While reference‐based deconvolution approaches offer robust solutions, they rely heavily on the availability and quality of external single‐cell reference data, which may be incomplete, unavailable, or poorly matched to the spatial data. Moreover, even when such references are available, they often represent only broad cell types, potentially obscuring finer subpopulation structures and masking intra‐type heterogeneity. To overcome these limitations, we introduced gnSPADE, a reference‐free spatial deconvolution method that incorporates gene network structures via a Markov random field within a latent Dirichlet allocation modeling framework. gnSPADE jointly infers cell type‐specific transcriptional profiles and spatial compositions without external references. When applied to synthetic and real ST datasets, gnSPADE achieves improved accuracy, spatial resolution, and biological interpretability compared to other methods, highlighting the power of reference‐free deconvolution in resolving complex tissues.

Year:

2026

Cancer is a highly heterogeneous disease characterized by complex molecular changes. Subtypes identified through multi-omics data hold significant promise for improving prognosis and facilitating personalized precision treatment. Recent multi-omics integration methods have mostly focused on capturing complementary information from different data types, often overlooking potential interactions between omics data. Here we develop a novel method named interactive multi-kernel learning (iMKL), which incorporates omics-omics interactions alongside heterogeneous data types under the unsupervised multi-kernel learning framework, to improve subtype identification. Using the sample-similarity kernel for each dataset, we propose a joint Hadamard product strategy to capture higher-order interactive effects from different omics data types. We applied iMKL to two renal cell carcinoma (RCC) datasets—clear renal cell carcinoma (ccRCC) and type II papillary renal cell carcinoma (type II pRCC)—both including miRNA expression, mRNA expression, and DNA methylation data. Stability analysis through random sampling of patients or features demonstrated that iMKL exhibits strong robustness and accuracy in identifying patient subtypes. The identified subtypes revealed dramatic differences in patient survival, with both ccRCC and type II pRCC classified into three distinct subtypes. The findings in the real application highlight potential biomarkers associated with adverse patient outcomes and demonstrate substantial advancement in cancer subtype identification. The iMKL method effectively identifies tumor molecular subtypes that are strongly associated with clinical features and survival rates, providing valuable insights for accurate cancer subtyping, clinical decision-making, and the realization of personalized treatment strategies.

Year:

2025

Objectives: Cancer, characterized by its profound complexity and heterogeneity, arises from a multitude of molecular disruptions. The pursuit of identifying distinct cancer subtypes is driven by the need to stratify patients into clinically coherent subgroups, each exhibiting unique prognostic outcomes. The integration of multi-omics datasets enhances the precision of subtyping and advances precision medicine. Methods: Considering the high-dimensional nature inherent to various multi-omics data types, we introduce an innovative deep learning framework, DAE-MKL, which integrates denoising autoencoders with multi-kernel learning for identifying cancer subtypes. Leveraging the capabilities of DAE, we extract non-linearly transformed features that retain pertinent information while mitigating noise and redundancy. These refined data representations are then funneled into the MKL framework, thereby enhancing the accuracy of subtype identification. We applied the DAE-MKL framework to both simulated studies and empirical datasets derived from two distinct cancer types, low-grade glioma (LGG, n = 86) and kidney renal clear cell carcinoma (KIRC, n = 285), thereby validating its utility and feasibility. Results: In simulations, DAE-MKL achieved superior performance with NMI gains up to 0.78 compared to other state-of-the-art methods. For real datasets, DAE-MKL identified three LGG subtypes and three KIRC subtypes, showing significant survival differences (KIRC log-rank p = 3.33 × 10−8, LGG log-rank p = 3.99 × 10−8). Additionally, we explored potential cancer-related biomarkers. Conclusions: The DAE-MKL effectively identifies molecular subtypes, reduces data dimensionality, and improves prognostic stratification in multi-omics cancer datasets, providing an effective tool for precision oncology.

Year:

2025

Most spatial transcriptomics (ST) technologies (e.g. 10× Visium) operate at the multicellular level, where each spatial location often contains a mixture of cells with heterogeneous cell types. Thus, effective deconvolution of cell type compositions is critical for downstream analysis. Although reference-based deconvolution methods have been proposed, they depend on the availability of reference data, which may not always be accessible. Additionally, within a deconvolved cell type, cellular heterogeneity may still exist, requiring further deconvolution to uncover finer structures for a better understanding of this complexity. Here, we present gwSPADE, a gene frequency-weighted reference-free SPAtial DEconvolution method for ST data. gwSPADE requires only the gene count matrix and utilizes appropriate weighting schemes within a topic model to accurately recover cell type transcriptional profiles and their proportions at each spatial location, without relying on external single-cell reference information. In various simulations and real data analyses, gwSPADE demonstrates scalability across various platforms and shows superior performance over existing reference-free deconvolution methods such as STdeconvolve.

Year:

2025

Recent advances in spatial transcriptomics (ST) have significantly deepened our understanding of biology. A primary focus in ST analysis is to identify spatially variable genes (SVGs) which are crucial for downstream tasks like spatial domain detection. Spatial domains reflect underlying tissue architecture and distinct biological processes. Traditional methods often use a set number of top SVGs for this purpose, and embedding these SVGs simultaneously can confound unrelated spatial signals, dilute weaker patterns, leading to obscured latent structure. Instead, grouping SVGs and getting low-dimensional embedding within each group preserves specific patterns, reduces signal mixing, and enhances the detection of diverse structures. Furthermore, classifying SVGs is akin to identifying cell-type marker genes, offering valuable biological insights. The challenge lies in accurately categorizing SVGs into relevant clusters, aggravated by the absence of prior knowledge regarding the number and spatial gene patterns. Here, we propose SPACE, a framework that classifies SVGs based on their spatial patterns by adjusting for shared cell-type confounding effects, to improve spatial domain detection. This method does not require prior knowledge of gene cluster numbers, spatial patterns, or cell type information. Both simulation and real data analyses demonstrate that SPACE is an efficient and promising tool for ST analysis.

Year:

2025

Spatial transcriptomics (ST) measures gene expression while preserving spatial context within tissues. One of the key tasks in ST analysis is spatial domain detection, which remains challenging due to the complex structure of ST data and the varying performance of individual clustering methods. To address this, we propose SpatialESD, a Spatial EnSemble Domain detection method that integrates results from different spatial domain detection methods to improve spatial domain detection. SpatialESD captures both direct co-occurrence patterns and multiscale indirect relationships between clusters, improving the robustness and accuracy of spatial domain detection. We evaluated SpatialESD on simulated datasets and multiple 10x Visium spatial transcriptomics datasets, including human brain, breast cancer, and ovarian cancer samples. The results show that SpatialESD consistently outperforms individual methods and the existing EnSDD ensemble method in terms of clustering accuracy and stability. Based on the identified domains, we further detected region-specific differentially expressed genes and performed trajectory and cell-cell interaction analyses. These results reveal spatial patterns of gene expression and cellular communication, offering insights into tissue organization and disease mechanisms. Overall, SpatialESD provides a reliable and effective solution for spatial domain detection in ST data and facilitates downstream biological discovery. Graphical Abstract

Year:

2025

Identifying spatially variable genes (SVGs) has been an essential task in spatial transcriptomics. In addition to SVGs detection, there are genes exhibiting expression patterns that are associated with cellular developmental stages or lineage fates across a tissue section. Identifying such genes could provide novel insights into tumor metastasis. Here, we introduce a unified statistical model, termed TrajEctory-preServed and SpAtially-variable gene detection (TESSA), to detect both types of genes. Moreover, we propose a novel strategy to address the inherent double-dipping issue commonly encountered when assessing temporal gene effects in transcriptomics analysis. We demonstrate the testing performance through extensive simulation studies and real applications to several publicly available datasets. Downstream analyses further highlight the potential of our method in identifying genes associated with tumor progression and enhancing spatial domain detection.

Year:

2025

Spatial transcriptomics (ST) technologies have offered unprecedented insights into the spatial organization of gene expression, allowing for the study of tissue architecture, domain boundaries, and cell-cell interactions. However, most ST data generated so far are at multicellular resolution, where each spot captures transcripts from a mixture of diverse cells of different cell types. While reference-based deconvolution approaches offer robust solutions, they rely heavily on the availability and quality of external single-cell reference data, which may be incomplete, unavailable, or poorly matched to the spatial data. Moreover, even when such references are available, they often represent only broad cell types, potentially obscuring finer subpopulation structures and masking intra-type heterogeneity. To overcome these limitations, we introduce gnSPADE, a reference-free spatial deconvolution method that incorporates gene network structures via a Markov random field within a latent Dirichlet allocation (LDA) modeling framework. gnSPADE jointly infers cell type-specific transcriptional profiles and spatial compositions without external references. Applied to synthetic and real ST datasets, gnSPADE achieves improved accuracy, spatial resolution, and biological interpretability compared to other methods, highlighting the power of reference-free deconvolution in resolving complex tissues.

Year:

2025

Year:

2025

Year:

2025

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2022

Collaborators (5)

yuesong wu

Michigan State University

UNITED STATES
View details

Yuying Xie

Associate Professor

Michigan State University

UNITED STATES
View details

Ling Wang

Associate Professor

Michigan State University

UNITED STATES
View details

Bin Chen

Assistant/Associate Professor

Michigan State University

UNITED STATES
View details

Nina Steele

Assistant Professor

University of Cincinnati

UNITED STATES
View details