Spatial Proteomics vs. Single-Cell Proteomics: Choosing the Right Approach

Researchers studying biological systems face a fundamental choice when designing proteomics experiments: whether to measure proteins across many individual cells dissociated from tissue, or to measure proteins within their native tissue architecture. Single-cell proteomics approaches such as mass cytometry and CITE-seq provide high-throughput protein measurements from individual cells but destroy spatial context during sample preparation. Spatial proteomics methods, including highly multiplexed tissue imaging and mass spectrometry imaging, preserve the tissue microenvironment and reveal how protein expression relates to cellular neighborhoods, tissue architecture, and cell-cell interactions. This article compares these two approach families across resolution, multiplexing capacity, throughput, and biological question suitability, and provides a decision framework for selecting the appropriate method for a given research goal.

Defining the Two Approach Families

Single-cell proteomics encompasses techniques that measure protein expression in individual cells after tissue dissociation. Mass cytometry uses metal-tagged antibodies to quantify dozens of proteins per cell at high throughput. CITE-seq combines antibody-derived tags with single-cell RNA sequencing to simultaneously measure surface proteins and transcriptomes. These methods excel at characterizing cellular heterogeneity across large cell populations and identifying rare cell states, but they require cells to be removed from their tissue context.

Spatial proteomics preserves tissue architecture and measures protein expression with positional information. Highly multiplexed tissue imaging platforms use DNA oligonucleotide-tagged, fluorophore-tagged, or metal-tagged antibodies to capture tens to over 100 markers within the same tissue specimen. These approaches generate spatially indexed datasets that reveal cell neighborhoods, spatial programs, and interaction patterns. Mass spectrometry-based spatial proteomics methods, including matrix-assisted laser desorption ionization imaging and liquid chromatography-tandem mass spectrometry approaches, provide deeper proteome coverage but typically at lower multiplexing or spatial resolution.

The distinction matters for experimental design because each approach answers different biological questions. Single-cell methods ask what cell types and states exist in a sample. Spatial methods ask where those cells are located, what their neighbors are, and how tissue organization relates to function or disease outcome.

Core Principles of Single-Cell Proteomics

Single-cell proteomics directly measures proteins and proteoforms in individual cells, providing information that transcriptomic methods cannot capture because most biological processes are controlled at the protein level. The field has advanced through developments in sample preparation, instrumentation, and informatics, yet it remains difficult and slow relative to competing single-cell technologies. What single-cell proteomics may lose in relative throughput, it trades for direct analysis of proteins, albeit with biases toward those of the highest relative concentration in each cell.

Mass cytometry, also called CyTOF, uses antibodies conjugated to rare earth metals. Cells are stained with the antibody panel, nebulized into single droplets, and introduced into a mass spectrometer that quantifies each metal tag. This approach routinely measures 40 to 50 proteins per cell across millions of cells per experiment. The technology borrows heavily from flow cytometry workflows, and many analysis tools developed for flow cytometry can be adapted for mass cytometry data.

CITE-seq and related approaches use oligonucleotide-tagged antibodies that are read out through single-cell RNA sequencing. These methods provide simultaneous measurement of surface protein expression and the transcriptome from the same cell. The protein panel is typically limited to 100 to 200 antibodies, but the transcriptome coverage provides genome-wide context. This dual measurement enables researchers to link protein phenotypes to transcriptional states and regulatory programs.

The primary limitation of all single-cell proteomics methods is the loss of spatial information. Tissue must be dissociated into a single-cell suspension, which destroys the architecture that defines tissue function. Cell-cell interactions, neighborhood composition, and tissue-level organization are lost during sample preparation. Additionally, dissociation itself can alter protein expression through enzymatic stress and loss of surface epitopes.

Core Principles of Spatial Proteomics

Spatial proteomics technologies have transformed understanding of complex tissue architecture in cancer and other diseases. These methods measure protein expression in situ, preserving the tissue context that determines cellular function. The field encompasses three main technical categories: antibody-based imaging, liquid chromatography-tandem mass spectrometry-based approaches, and imaging mass spectrometry.

Antibody-based spatial proteomics platforms use multiplexed immunohistochemistry or immunofluorescence to visualize dozens of proteins in a single tissue section. Highly multiplexed tissue imaging platforms leverage antibody labeling strategies including DNA oligonucleotide-tagged, fluorophore-tagged, and metal-tagged reagents to capture tens to over 100 markers within the same specimen. These approaches generate rich spatially indexed datasets that can reveal cell neighborhoods, spatial programs, and interaction patterns.

Imaging mass spectrometry methods, including matrix-assisted laser desorption ionization, measure proteins directly from tissue sections without antibody labeling. These approaches provide unbiased detection of hundreds to thousands of proteins but at lower spatial resolution and throughput than antibody-based methods. The non-amplifiable nature of proteins and the sensitivity limitations of mass spectrometry create challenges for achieving bulk tissue-level coverage with high spatial resolution.

Recent advances have pushed spatial proteomics toward deeper coverage and higher throughput. One strategy uses computationally assisted image reconstruction with sparse sampling to reduce the number of mass spectrometry measurements needed for whole-tissue mapping. This approach generated the largest spatial proteome to date, mapping more than 9000 proteins in the mouse brain, and discovered potential new regional or cell type markers. Another approach uses tissue expansion to achieve micrometre-resolution deep spatial proteomics.

At a Glance: Comparison Table

Resolution and Coverage Tradeoffs

Spatial resolution and proteome coverage exist in tension across all proteomics platforms. Single-cell proteomics methods achieve single-cell resolution by definition, but they sacrifice the tissue context that defines cellular function. Imaging-based spatial proteomics achieves single-cell resolution while preserving tissue architecture, but the number of proteins that can be measured simultaneously is limited by antibody panel design and spectral or mass overlap.

Mass spectrometry-based spatial proteomics provides the deepest proteome coverage but faces challenges in spatial resolution and throughput. The non-amplifiable nature of proteins means that sensitivity cannot be improved through amplification steps common in genomics. Limited multiplexing capability means that whole-tissue slice mapping with high spatial resolution requires a formidable amount of mass spectrometry matching time. Sparse sampling strategies address this limitation by reducing the number of samples needed by tens to thousands of times depending on the desired spatial resolution.

The choice between resolution and coverage depends on the biological question. Studies focused on cellular phenotypes and rare cell states benefit from the deep single-cell coverage of mass cytometry or CITE-seq. Studies focused on tissue organization and microenvironment interactions require spatial methods even if the marker panel is more limited. Studies seeking unbiased discovery of regional protein markers benefit from mass spectrometry-based spatial proteomics despite lower throughput.

Multiplexing Capacity and Panel Design

Multiplexing capacity determines how many proteins can be measured in a single experiment. Single-cell proteomics platforms vary widely in this regard. Mass cytometry panels typically include 40 to 50 antibodies, limited by the availability of rare earth metal isotopes and the need to avoid signal overlap. CITE-seq panels can include 100 to 200 antibodies, with the oligonucleotide tags read out through sequencing. Mass spectrometry-based single-cell proteomics can detect hundreds to thousands of proteins, but with biases toward highly abundant proteins and lower throughput.

Highly multiplexed tissue imaging platforms have expanded multiplexing capacity through innovative labeling strategies. DNA oligonucleotide-tagged antibodies enable iterative staining and imaging cycles, allowing tens to over 100 markers to be measured in the same specimen. Fluorophore-tagged approaches use spectral unmixing to distinguish multiple signals in a single imaging round. Metal-tagged approaches use mass cytometry technology adapted for imaging, with each pixel representing a mass spectrum.

Panel design requires careful consideration of antibody specificity, signal dynamic range, and potential spillover between channels. The rapidly expanding landscape of highly multiplexed tissue imaging creates practical challenges in platform selection, computational analysis, and cross-study reproducibility. Each study uses a different marker panel and protocol, and most methods are tailored to single cohorts, which limits knowledge transfer and robust biomarker discovery.

Workflow Considerations for Each Approach

Single-Cell Proteomics Workflow

The single-cell proteomics workflow begins with tissue collection and dissociation. Fresh tissue must be processed quickly to maintain cell viability and preserve surface epitopes. Enzymatic digestion and mechanical disruption generate a single-cell suspension, but these steps can alter protein expression and introduce artifacts. Cells are then stained with the antibody panel, washed, and analyzed on the instrument.

Mass cytometry requires cells to be fixed and often permeabilized before staining. The stained cells are introduced into the instrument as a single-cell stream, and each cell generates a mass spectrum that is converted to protein expression values. Data analysis follows flow cytometry conventions, including gating, compensation, and dimensionality reduction approaches such as t-distributed stochastic neighbor embedding and uniform manifold approximation and projection.

CITE-seq requires cells to be encapsulated in droplets or sorted into wells for sequencing library preparation. The antibody-derived tags are amplified and sequenced alongside the transcriptome. Data analysis requires separate processing of the protein and RNA components, followed by integrated analysis to identify cell types and states.

Spatial Proteomics Workflow

Spatial proteomics workflows begin with tissue sectioning and mounting onto slides. Frozen tissue sections are commonly used for mass spectrometry-based methods, while formalin-fixed paraffin-embedded sections can be used for antibody-based imaging after antigen retrieval. The choice of tissue preparation affects antibody compatibility and protein detection sensitivity.

Antibody-based imaging workflows involve staining the tissue section with the antibody panel, imaging the section, and processing the images to extract single-cell data. Cell segmentation identifies individual cells based on nuclear and membrane markers, and protein expression is quantified for each cell. The resulting data can be analyzed using flow cytometry-like workflows adapted for spatial data, with the key difference that spatial information is conserved at single-cell resolution.

Mass spectrometry-based spatial proteomics workflows involve applying a matrix to the tissue section and acquiring mass spectra at defined spatial positions. The resulting data are processed to generate ion images that map protein distributions across the tissue. Sparse sampling strategies reduce acquisition time by measuring only a subset of positions and using computational reconstruction to generate full tissue maps.

Data Analysis and Computational Requirements

Both approach families generate large, complex datasets that require specialized computational analysis. Single-cell proteomics data analysis has benefited from the mature ecosystem of flow cytometry and single-cell genomics tools. Mass cytometry data can be analyzed with established workflows including manual gating, automated clustering, and dimensionality reduction. CITE-seq data require integrated analysis of protein and RNA modalities, with tools developed for multi-modal single-cell data.

Spatial proteomics data analysis presents additional challenges because the data include positional information that must be incorporated into the analysis. Major analytical bottlenecks include preprocessing and normalization, denoising, spillover correction, cell segmentation, cell-type annotation, and tissue microarchitecture analysis. Method limitations, scalability, and performance under challenging conditions must be considered when selecting analysis tools.

Image cytometry approaches adapt flow cytometry workflows to spatial data while conserving spatial information at single-cell resolution. Spatial uniform manifold approximation and projection can be constructed using image cytometry output, enabling visualization of cell phenotypes in spatial context. Spatial UMAP subtraction analysis can identify topographic and coexpression signatures associated with clinical outcomes, highlighting immune neighborhoods and associated topographic immunoactive protein expression patterns.

Artificial intelligence approaches are emerging for spatial proteomics analysis. Foundation models that learn marker-aware, multi-scale representations of proteins, cells, niches, and tissues directly from multiplex imaging data can support marker reconstruction, cell typing, niche annotation, spatial biomarker discovery, and patient stratification. These models enable zero-shot annotation across heterogeneous panels and datasets, addressing the challenge of cross-study comparability.

Choosing Based on Research Question

The choice between spatial and single-cell proteomics should be driven by the biological question instead of technological availability. Studies of cellular heterogeneity benefit from single-cell approaches that profile large numbers of cells. Studies of tissue organization require spatial approaches that preserve architecture.

Research questions about cell type composition and frequency are well served by single-cell proteomics. Mass cytometry can profile millions of cells and identify rare populations that might be missed by imaging approaches limited to smaller tissue areas. CITE-seq adds transcriptome information that can reveal regulatory programs underlying protein phenotypes.

Research questions about tissue architecture and microenvironment organization require spatial proteomics. The tumor microenvironment significantly influences clinical response to immune checkpoint inhibition, and spatial mapping of protein distributions in tumor-infiltrating lymphocyte enriched versus low compartments can identify targetable biologic processes. Spatial proteomics revealed that sirtuin 1 was enriched in CD8-high tumor compartments, and pharmacological activation of sirtuin 1 increased the immunological effect of anti-PD-1 immune checkpoint inhibition in melanoma mouse models.

Research questions about cell-cell interactions and neighborhood composition are uniquely addressed by spatial methods. Multiplex immunofluorescence can detail spatial relationships and complex cell phenotypes in the tumor microenvironment. Analysis of patient biopsies via spatial proteomics revealed that in responders to checkpoint inhibition, there was increased B cell activation, mature tertiary lymphoid structures, and increased CD8+ T cell-macrophage distances with treatment.

Decision Framework for Method Selection

The following decision framework guides method selection based on research goals, sample availability, and analytical resources.

Step 1: Define the Primary Biological Question

State whether the question concerns cellular phenotypes in isolation or cellular phenotypes in tissue context. Questions about cell-intrinsic states, such as signaling pathway activation or differentiation status, can be addressed with single-cell methods. Questions about tissue-level organization, such as immune cell infiltration patterns or tumor-stroma interactions, require spatial methods.

Step 2: Assess Sample Availability and Characteristics

Determine whether fresh tissue can be obtained for dissociation or whether only fixed tissue sections are available. Single-cell proteomics requires fresh or optimally preserved tissue for dissociation. Spatial proteomics can work with formalin-fixed paraffin-embedded tissue for antibody-based methods, while mass spectrometry-based methods typically require frozen sections.

Step 3: Evaluate Required Multiplexing and Coverage

Consider how many proteins must be measured to answer the research question. Hypothesis-driven studies with defined marker panels can use imaging-based spatial proteomics or mass cytometry. Discovery-oriented studies seeking unbiased proteome coverage benefit from mass spectrometry-based approaches, either single-cell or spatial.

Step 4: Consider Throughput and Scale

Estimate the number of samples and cells that must be analyzed. Single-cell methods profile millions of cells per experiment, enabling population-level statistics. Spatial methods profile thousands to hundreds of thousands of cells per tissue section, with imaging time and field size limiting throughput.

Step 5: Assess Computational Resources and Expertise

Evaluate the analytical pipeline required for each approach. Single-cell proteomics benefits from mature analysis tools but requires expertise in high-dimensional data analysis. Spatial proteomics requires additional computational steps for image processing, cell segmentation, and spatial analysis, with emerging artificial intelligence tools addressing some of these challenges.

Step 6: Plan for Validation and Integration

Consider whether the chosen approach will be validated with orthogonal methods. Multi-modal strategies that leverage the strengths of each platform can identify tumor-specific features and microenvironmental patterns that might be missed by any single approach. Integration with transcriptomic, genomic, and clinical data can provide context for proteomic findings.

Records and Measurements for Quality Control

Quality control records should document every step of the experimental workflow to ensure reproducibility and enable troubleshooting. For single-cell proteomics, records should include tissue collection time, dissociation protocol details, cell viability measurements, antibody panel lot numbers, instrument settings, and data acquisition parameters. For spatial proteomics, records should include tissue fixation and embedding details, section thickness, antigen retrieval conditions, antibody validation data, imaging parameters, and segmentation quality metrics.

Cell density and proportion measurements from image cytometry showed strong correlations with gold-standard digital pathology software, with correlation coefficients above 0.8, indicating that image-based quantification can match established methods when properly validated. Quality control should include comparison with orthogonal methods where possible.

For mass spectrometry-based approaches, records should include protein extraction efficiency, digestion completeness, chromatography performance, mass spectrometer calibration, and data-dependent acquisition settings. Sparse sampling strategies require documentation of sampling density and reconstruction parameters.

Common Failure Patterns and Troubleshooting

Several failure patterns recur across proteomics experiments. Antibody panels that have not been validated for the specific tissue type or fixation method produce unreliable measurements. Single-cell dissociation protocols that are too harsh degrade surface epitopes and alter protein expression. Imaging experiments with poor segmentation produce inaccurate single-cell measurements. Mass spectrometry experiments with insufficient sensitivity miss low-abundance proteins.

Cross-study reproducibility remains a major challenge in spatial proteomics. Each study uses a different marker panel and protocol, and most methods are tailored to single cohorts, which limits knowledge transfer and robust biomarker discovery. Standardization of panels, protocols, and analysis pipelines is needed to enable meta-analysis across studies.

Spillover between detection channels can confound measurements in both mass cytometry and multiplexed imaging. Careful panel design and compensation are required to minimize these effects. Denoising and spillover correction are essential preprocessing steps for highly multiplexed tissue imaging data.

Cell segmentation errors propagate through the entire analysis pipeline. Poor segmentation can merge adjacent cells or split single cells, producing inaccurate protein expression measurements and cell type assignments. Quality control should include visual inspection of segmentation results and comparison with manual annotation for a subset of images.

Limitations and Interpretation Boundaries

Single-cell proteomics methods have inherent biases toward highly abundant proteins. Mass spectrometry-based single-cell proteomics detects proteins with biases toward those of the highest relative concentration in each cell, potentially missing biologically important low-abundance proteins. Antibody-based methods are limited by antibody availability and specificity, and the panel design constrains the biological questions that can be addressed.

Spatial proteomics methods face different limitations. Antibody-based imaging is limited to tens to over 100 markers, which may not capture the full complexity of cellular states. Mass spectrometry-based spatial proteomics has lower sensitivity and throughput, limiting its application to abundant proteins and smaller tissue areas. The non-amplifiable nature of proteins means that sensitivity cannot be improved through amplification, and limited multiplexing capability requires tradeoffs between spatial resolution and coverage.

Interpretation of spatial proteomics data requires careful consideration of the tissue context. Protein expression patterns may reflect cellular composition differences instead of changes in protein expression per cell. Cell density and neighborhood composition must be accounted for when interpreting spatial patterns. Spatial UMAP analysis can identify immune neighborhoods and associated topographic immunoactive protein expression patterns, but these patterns require validation with supervised approaches.

Safety and Regulatory Context

Proteomics research involving human tissue samples must comply with institutional review board requirements and applicable regulations. Data sharing must follow established policies for genomic and related data. The National Institutes of Health Genomic Data Sharing Policy provides requirements for data sharing that may apply to proteomics data generated from human subjects. Researchers should consult their institution's data sharing policies and the funding agency requirements before initiating studies.

Data management should follow the FAIR Guiding Principles, which emphasize findability, accessibility, interoperability, and reusability of data. These principles provide a framework for making proteomics data maximally useful to the research community. Data repositories such as those maintained by the National Center for Biotechnology Information and the European Bioinformatics Institute provide infrastructure for data deposition and access.

Training resources are available through organizations such as the European Bioinformatics Institute, which offers courses and materials on bioinformatics topics including proteomics data analysis. Researchers should ensure that personnel have appropriate training before conducting complex proteomics experiments.

Professional Escalation Criteria

Researchers should seek expert consultation when facing specific challenges in proteomics experiment design or analysis. Consult a bioinformatics specialist when the computational requirements exceed local expertise, particularly for spatial proteomics data analysis involving cell segmentation, spatial statistics, or artificial intelligence approaches. Consult a mass spectrometry specialist when sensitivity or throughput limitations prevent achieving the required proteome coverage. Consult a statistics specialist when designing studies with multiple comparisons or when interpreting complex spatial patterns.

Escalate to institutional review or regulatory bodies when the research involves human subjects, protected health information, or data sharing requirements that raise compliance questions. Escalate to clinical collaborators when findings have potential diagnostic or therapeutic implications that require clinical validation.

Frequently Asked Questions

What is the main difference between spatial proteomics and single-cell proteomics?

Spatial proteomics preserves the tissue architecture and measures proteins in their native context, revealing where proteins are expressed and how cells interact with their neighbors. Single-cell proteomics dissociates tissue into individual cells and measures proteins in each cell, providing high-throughput characterization of cellular heterogeneity but losing all positional information.

When should I choose single-cell proteomics over spatial proteomics?

Choose single-cell proteomics when the research question concerns cellular phenotypes in isolation, such as identifying cell types, characterizing differentiation states, or profiling rare populations. Single-cell methods profile millions of cells per experiment and are well suited for population-level statistics and discovery of rare cell states.

When should I choose spatial proteomics over single-cell proteomics?

Choose spatial proteomics when the research question concerns tissue organization, cell-cell interactions, or the tumor microenvironment. Spatial methods reveal cell neighborhoods, spatial programs, and interaction patterns that cannot be captured after tissue dissociation. Spatial proteomics has identified microenvironmental patterns enriched in high-risk patients and revealed proteins that determine tumor-infiltrating lymphocyte infiltration.

Can I combine spatial and single-cell proteomics in one study?

Yes, multi-modal strategies that leverage the strengths of each platform can provide complementary information. Combining single-cell RNA sequencing with spatial proteomics can link transcriptional states to spatial organization. Multi-omics approaches have identified glycolysis-associated cellular communities and their cancer-promoting mechanisms by integrating single-cell and spatial data.

What are the throughput limitations of spatial proteomics?

Imaging-based spatial proteomics is limited by imaging time and field size, typically profiling thousands to hundreds of thousands of cells per tissue section. Mass spectrometry-based spatial proteomics has lower throughput due to acquisition time requirements. Sparse sampling strategies can reduce the number of samples needed by tens to thousands of times depending on the desired spatial resolution.

How many proteins can be measured with each approach?

Mass cytometry typically measures 40 to 50 proteins per cell. CITE-seq can measure 100 to 200 surface proteins alongside the transcriptome. Highly multiplexed tissue imaging can capture tens to over 100 markers in the same specimen. Mass spectrometry-based spatial proteomics can detect hundreds to thousands of proteins, with recent approaches mapping more than 9000 proteins in the mouse brain.

What are the main challenges in spatial proteomics data analysis?

Major analytical bottlenecks include preprocessing and normalization, denoising, spillover correction, cell segmentation, cell-type annotation, and tissue microarchitecture analysis. Each study uses a different marker panel and protocol, which limits knowledge transfer and robust biomarker discovery. Artificial intelligence approaches are emerging to address these challenges through foundation models that learn marker-aware representations directly from multiplex imaging data.

How do I validate spatial proteomics findings?

Validation should include orthogonal methods such as multiplexed immunohistochemistry, western blotting, or immunofluorescence. Cell densities and proportions identified by image cytometry showed strong correlations with gold-standard digital pathology software. Functional validation in model systems can confirm the biological relevance of spatial proteomics findings, as demonstrated by pharmacological modulation of sirtuin 1 activity in syngeneic mouse models.

Related Bioinformatics Guides

• Modern Transcriptomics: From Bulk RNA-Seq to Single-Cell and Spatial Resolution
• Single-Cell ATAC-Seq Bioinformatics
• Single-Cell RNA Sequencing: From Bulk to Resolution
• Single-Cell RNA-Seq Analysis Pipelines for Veterinary Immunology
• Single-cell RNA-seq Trajectory Inference and Cell Lineage Tracing

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Multi-omic profiling of follicular lymphoma reveals changes in tissue architecture and enhanced stromal remodeling in high-risk patients.. Cancer cell, 2024.
• Spatial proteomics reveals sirtuin 1 to be a determinant of T-cell infiltration in human melanoma.. The British journal of dermatology, 2025.
• Multi-omics-driven discovery of invasive patterns and treatment strategies in CA19-9 positive intrahepatic cholangiocarcinoma.. Journal of translational medicine, 2024.
• AI-driven multimodal algorithm predicts immunotherapy and targeted therapy outcomes in clear cell renal cell carcinoma.. Cell reports. Medicine, 2025.
• Spatial TCR clonality and clonal expansion in the in situ microenvironment of non-small cell lung cancer.. Journal for immunotherapy of cancer, 2025.
• Pooled tagging and hydrophobic targeting of endogenous proteins for unbiased mapping of unfolded protein responses.. Molecular cell, 2025.
• Spatial UMAP and Image Cytometry for Topographic Immuno-oncology Biomarker Discovery.. Cancer immunology research, 2021.
• Cancer systems immunology reveals myeloid-T cell interactions and B cell activation mediate response to checkpoint inhibition in metastatic breast cancer.. bioRxiv : the preprint server for biology, 2025.
• Spotlight on challenges and novel methods in highly multiplexed tissue imaging-based spatial proteomics.. 2026.
• What Are the Practical Applications of Single-Cell Proteomics?. 2026.
• Urinary Proteomics Identifies ADAMDEC1 as a Non-Invasive Biomarker of Unstable Carotid Atherosclerotic Plaques.. 2026.
• The evolving landscape of spatial proteomics technologies in the AI age. Fundamental Research, 2024.
• AI-powered virtual tissues from spatial proteomics for clinical diagnostics and biomedical discovery. arXiv.org, 2025.
• In-depth and high-throughput spatial proteomics for whole-tissue slice profiling by deep learning-facilitated sparse sampling strategy. Cell Discovery, 2025.
• iPEX enables micrometre-resolution deep spatial proteomics via tissue expansion. Nature, 2025.
• Spatial proteomics for investigating solid tumor resistance mechanisms. Cancer Metastasis Review, 2025.
• Single-cell and spatial transcriptomic analysis reveals that an immune cell-related signature could predict clinical outcomes for microsatellite-stable colorectal cancer patients receiving immunotherapy. Journal of Zhejiang University Science B, 2025.
• Multiscale, multimodal analysis of tumor heterogeneity in IDH1 mutant vs wild-type diffuse gliomas. Plos One, 2019.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.