MALDI-MSI and Visium Spatial Transcriptomics Integration: Best-Practice Workflows for Oncology and Neurodegeneration Biomarker Discovery
Best-practice workflows for integrating MALDI mass spectrometry imaging with Visium spatial transcriptomics, covering same-section and adjacent-section strategies, computational registration tools, and mechanistic validation.
Advanced Experimental Methods
MALDI-MSI and Visium Spatial Transcriptomics Integration: Best-Practice Workflows for Oncology and Neurodegeneration Biomarker Discovery
Integrating MALDI mass spectrometry imaging with Visium spatial transcriptomics on the same or adjacent tissue sections enables simultaneous mapping of metabolite distributions and gene expression at cellular resolution. Computational tools including ESCDAT, MIIT, and SpatialData now support single-section pixel-scale alignment and non-rigid serial-section registration, while proximity-based labeling with TurboID provides the mechanistic grounding needed to convert spatial discoveries into validated biomarkers.
What it does
MALDI-MSI and spatial transcriptomics integration co-localizes label-free metabolite and lipid maps with genome-wide or targeted transcript measurements across intact tissue architecture. MALDI-MSI acquires ion images at 5 to 50 µm lateral resolution by rastering a pulsed laser across a matrix-coated tissue section, desorbing and ionizing analytes for time-of-flight or Orbitrap mass analysis. Spatial transcriptomics platforms such as Visium (55 µm spot capture arrays) or Xenium (subcellular in situ decoding) provide complementary transcriptomic layers on the same or serial sections. Computational registration fuses both modalities into a unified spatial coordinate system, exposing metabolic-genomic coupling at the level of individual cells or defined tissue compartments.
Why use this method
Transcriptomics alone captures gene regulation but not the downstream metabolic or lipid state of a cell. A tumor cell overexpressing fatty acid synthase may or may not accumulate specific phosphatidylcholines depending on post-transcriptional and enzymatic constraints that RNA cannot reveal. MALDI-MSI fills this gap by directly imaging the spatial distribution of hundreds of metabolites, lipids, and glycans in a single acquisition without requiring prior knowledge of target analytes.
Conversely, MALDI-MSI without transcriptomics lacks cell-type identity. A lipid enrichment in a tumor core could originate from cancer cells, tumor-associated macrophages, or cancer-associated fibroblasts. Pairing MALDI-MSI with spatially resolved transcriptomics assigns each ion signal to transcriptomically defined cell populations, converting metabolite maps from anatomical descriptions into cell-type-resolved biochemical phenotypes.
The combined method is particularly powerful for archival FFPE material, where established protocols now enable WES, RNA-seq, and metabolite-mode MALDI-MSI from a single section, with greater than 91% correlation between treated and untreated transcriptomes (PMID: 36045222). This compatibility with FFPE removes a major barrier to retrospective clinical cohort analysis.
Technical integration approaches
Same-section sequential analysis with ESCDAT
MALDI-MSI is performed first at 5 µm resolution on a 10 µm FFPE or cryo-section, creating a regular grid of fluorescence micro-dots through laser ablation. The section is then processed for single-cell spatial transcriptomics on the Xenium platform. ESCDAT (MATLAB-based) imports cell boundary coordinates from Xenium and uses the ablation micro-dots as internal fiducial markers to achieve pixel-scale co-registration. Per-cell mass spectra are extracted by assigning each MSI pixel to the overlapping cell boundary polygon, directly linking metabolic state to transcriptomic identity. This workflow introduces approximately 20 to 30% fewer transcript counts per cell due to ablation, but overall transcriptome complexity and cell-type classification accuracy remain preserved (PMID: 41315396).
Adjacent-section registration with MIIT and GreedyFHist
When sample preparation requirements make same-section analysis impractical, the Multi-Omics Imaging Integration Toolset (MIIT) registers MALDI-MSI and Visium ST data acquired from serial sections. The GreedyFHist non-rigid registration algorithm applies YOLO8-based background segmentation and center-of-mass preprocessing to align H&E or HES-stained histology anchors. Groupwise registration through intermediate sections handles heterogeneity changes across sections separated by up to 100 µm, achieving a median target registration error of 37.464 µm versus 216.054 µm for the HistoReg comparator. MIIT then fuses data using weighted statistics over the shared area between each Visium spot and the MSI pixels it contains. MIIT and GreedyFHist are open-source and Python-based, with QuPath integration for digital pathology pipelines (PMID: 40366868).
FAIR multimodal alignment with SpatialData
SpatialData provides an open-source Python framework using OME-NGFF and Zarr to store and align multimodal spatial datasets in a findable, accessible, interoperable, and reusable (FAIR) format. Coordinate transformations map Visium spots, Xenium transcripts, MALDI-MSI ion images, and H&E whole-slide images into a common coordinate system. SpatialData is modality-agnostic and supports downstream analysis in scanpy and squidpy, making it the preferred data container for projects combining three or more spatial modalities (PMID: 38509327).
Morphology-aware integration with MISO and OmiCLIP
MISO applies spectral clustering to jointly analyze omics layers and histomorphology, identifying fine-grained tissue structures such as high endothelial venules that are invisible to omics data alone (PMID: 39815104). OmiCLIP uses contrastive learning trained on over 2 million image-transcriptomics pairs to align H&E morphological features with genomic measurements, bridging the interpretation gap between pathologist annotation and molecular profiling (PMID: 40442373). Both tools extend single-modality MALDI-ST integration toward full multi-modal atlasing.
Prime applications
Oncology: glioblastoma lipid profiling and NSCLC survival prediction
Single-cell MALDI-MSI combined with immunohistochemistry (MALDI-IHC) on patient-derived glioblastoma cells resolves cell-type-specific lipid profiles at single-cell resolution. Phosphatidylcholine PC 36:4 is enriched specifically in tumor cells relative to microenvironmental populations, providing a spatially anchored lipid marker that distinguishes tumor cells without requiring antibody panels (PMID: 39932302).
In non-small cell lung cancer (NSCLC), spatial multi-omic signatures derived from integrated MALDI-MSI and transcriptomics predict progression-free survival outcomes, establishing spatial metabolic-genomic co-features as clinically actionable prognostic variables (PMID: 41073787).
Neurodegeneration: ganglioside mapping in Alzheimer's disease
MALDI-MSI co-registered with histology in Alzheimer's disease brain sections demonstrates specific co-localization of ganglioside GM3 and GM1 species with amyloid-beta plaques. A decrease in the GM1 d20:1/d18:1 ratio is observed specifically in the entorhinal cortex, one of the earliest affected regions in AD. These spatially resolved lipid signatures correlate amyloid plaque burden with ganglioside compositional shifts in a region- and isoform-specific manner (PMID: 38896306).
Archival FFPE cohort analysis
Protocols enabling simultaneous WES, RNA-seq, and metabolite-mode MALDI-MSI from a single FFPE section open retrospective biobanked cohorts to spatial multi-omics analysis. The high transcriptome correlation (greater than 91%) between MALDI-processed and unprocessed sections confirms that metabolite extraction does not substantively degrade genomic or transcriptomic data quality, making this approach suitable for clinical tissue archives where material is scarce (PMID: 36045222).
Validation strategy
Proximity-based interactome validation with TurboID and ProPPr
TurboID achieves promiscuous biotinylation of proximal proteins within 10 minutes in living systems, resolving transient and low-affinity interactors that co-immunoprecipitation misses (PMID: 30125270). For spatial proteomic discoveries in neurodegenerative tissue, the ProPPr (Probe-dependent Proximity Profiling) adaptation applies TurboID directly to human FFPE sections. Applied to phospho-tau in Alzheimer's disease and related tauopathies, ProPPr identified 1,317 phospho-tau-associated proteins, including specific sequestration of the retromer component VPS35 and the lysosomal membrane glycoprotein LAMP2 within pathological tau lesions (PMID: 40082954). These findings connect spatially observed protein co-localization to mechanistic pathway disruption in endo-lysosomal trafficking.
Endogenous validation via CRISPR knock-in of TurboID at the AP1M1 locus improves labeling specificity for transient interactors and clathrin-coated vesicle cargo relative to overexpression constructs, reducing background biotinylation from cytoplasmic TurboID pools (PMID: 39056144).
Correlative intravital microscopy for in vivo spatial confirmation
Three-photon microscopy (3PM) and related deep-tissue intravital approaches provide in vivo spatial ground truth for biomarkers identified by MALDI-MSI and spatial transcriptomics on fixed sections. In glioblastoma, correlative intravital imaging resolves tumor microtubes (TMs) and vascular invasion routes at subcellular resolution in live tumor tissue, providing a microscopic reference to validate clinical imaging markers such as diffusion tensor imaging (DTI) tractography. This direct comparison between ex vivo spatial omics findings and in vivo cellular behavior is the critical bridge from tissue atlas to mechanistic understanding (PMID: 39256378).
Functional and orthogonal omics validation
Candidate metabolite-transcript co-features identified in integrated MALDI-ST datasets require functional validation through perturbation experiments: genetic knockdown or knockout of enzymes upstream of the candidate metabolite, followed by MALDI-MSI to confirm ion-signal loss, and RNA-seq or scRNA-seq to confirm transcriptomic consequences. Orthogonal spatial validation uses immunofluorescence or multiplex protein imaging (e.g., CODEX, CyCIF) to confirm that protein-level expression of the implicated pathway matches the transcriptomic signature, closing the loop between metabolite imaging, gene expression, and protein abundance in the same tissue region.
Evidence quality and limitations
The evidence base for MALDI-MSI and spatial transcriptomics integration is grounded in multiple independent publications describing distinct tissue types, disease models, and instrument platforms. Same-section integration with ESCDAT is supported by direct experimental data from Xenium-MALDI co-acquisition on human tissue (PMID: 41315396). Adjacent-section MIIT registration is validated with quantified target registration error metrics across multiple tissue types and section distances (PMID: 40366868). SpatialData's FAIR framework has been adopted across multiple published spatial atlasing projects, providing a community-vetted data standard (PMID: 38509327). Clinical relevance is supported by disease-specific findings in glioblastoma, NSCLC, and Alzheimer's disease with named molecular targets (PC 36:4, VPS35, LAMP2, GM1, GM3), moving beyond proof-of-concept demonstrations toward biomarker-grade specificity.
Resolution mismatch remains the central technical limitation: MALDI-MSI typically achieves 5 to 50 µm lateral resolution, while Xenium subcellular transcriptomics achieves 0.2 µm resolution, creating a fundamental scale mismatch when assigning multi-pixel MSI spectra to single-cell boundaries. The same-section workflow incurs a 20 to 30% reduction in transcript counts per cell due to laser ablation, which may systematically bias detection of low-abundance transcripts. Adjacent-section registration introduces inter-section variability even at the best-reported AM-TRE of 37.464 µm, which exceeds the diameter of many mammalian cell types (10 to 20 µm) and makes single-cell correspondence across sections unreliable. MALDI-MSI analyte coverage is matrix-dependent: DHB matrices favor glycolipids and metabolites, while CHCA matrices favor peptides; no single matrix achieves comprehensive coverage. TurboID proximity labeling in FFPE tissue via ProPPr requires tissue that retains sufficient antigen accessibility after deparaffinization, and labeling radius (approximately 10 nm) captures only direct and very proximal interactors, potentially missing functional but distal pathway members.
The integration of MALDI-MSI with spatial transcriptomics represents a fundamental expansion of the spatial omics toolkit, enabling simultaneous metabolic and genomic phenotyping within intact tissue microenvironments. Computational frameworks including ESCDAT, MIIT, and SpatialData have brought this integration within reach of standard bioinformatics workflows, while proximity labeling with TurboID and deep-tissue intravital imaging provide the mechanistic grounding needed to move from spatial correlation to validated pathway disruption. The field is advancing rapidly from proof-of-concept studies toward clinical-grade spatial atlases of oncology and neurodegeneration tissue, with named molecular targets already emerging as candidate biomarkers across glioblastoma, NSCLC, and Alzheimer's disease. The primary remaining barriers are resolution harmonization across modalities, matrix-dependent analyte coverage in MALDI, and the need for larger prospective cohorts to establish the clinical predictive value of spatially derived multi-omic signatures.
Frequently asked questions
What is the key difference between same-section and adjacent-section MALDI-MSI and spatial transcriptomics integration?
Same-section integration, as implemented with ESCDAT, acquires MALDI-MSI at 5 µm resolution and then performs single-cell spatial transcriptomics on the identical 10 µm tissue section, yielding one-to-one cellular correspondence at the cost of approximately 20 to 30% fewer transcript counts per cell due to laser ablation. Adjacent-section integration, as implemented with MIIT and GreedyFHist, registers serial sections up to 100 µm apart using non-rigid algorithms anchored on H&E or HES staining, preserving full transcriptome complexity at the cost of introducing inter-section variability.
What computational tools support MALDI-MSI and spatial transcriptomics co-registration?
ESCDAT (MATLAB-based) uses MALDI laser-ablation fluorescence micro-dots as internal fiducials for pixel-scale alignment on the same section. MIIT (Python-based, open-source) uses GreedyFHist non-rigid registration with YOLO8 background segmentation for adjacent-section fusion. SpatialData provides a FAIR framework using OME-NGFF and Zarr to unify Visium, Xenium, and H&E data in common coordinate systems. MISO and OmiCLIP extend integration to morphology via spectral clustering and contrastive learning respectively.
How does TurboID proximity labeling validate spatial proteomic discoveries?
TurboID biotinylates interacting proteins within 10 minutes in living systems, enabling the identification of proximity-based interactomes with high temporal resolution. In FFPE neurodegenerative tissue, the ProPPr adaptation of TurboID identified 1,317 phospho-tau-associated proteins in Alzheimer's disease, including sequestration of VPS35 and LAMP2 in pathological lesions. CRISPR knock-in of TurboID at endogenous loci (e.g., AP1M1) further improves labeling specificity for transient interactors compared to overexpression models.
What lipid biomarkers has MALDI-MSI identified in glioblastoma?
Integrated single-cell MALDI-MSI and immunohistochemistry on patient-derived glioblastoma cells has identified cell-type-specific lipid profiles, including enrichment of phosphatidylcholine PC 36:4 in tumor cells. This spatial lipid profiling enables discrimination of tumor cell populations from microenvironmental cells at single-cell resolution.
What ganglioside changes has MALDI-MSI detected in Alzheimer's disease tissue?
MALDI-MSI co-registered with histology in Alzheimer's disease brain sections demonstrates specific co-localization of ganglioside GM3 and GM1 species with amyloid-beta plaques. A decrease in the GM1 d20:1/d18:1 ratio is observed in the entorhinal cortex, providing spatially resolved lipid signatures that correlate with amyloid pathology at a region- and isoform-specific level.
How does GreedyFHist non-rigid registration outperform affine methods for tissue alignment?
Affine registration applies only global transformations (translation, rotation, scaling, shear) and cannot compensate for local tissue deformations or artifacts in frozen sections. GreedyFHist uses YOLO8-based background segmentation, center-of-mass preprocessing, and groupwise registration to handle increasing section distances and tissue heterogeneity, achieving a median target registration error of 37.464 µm for adjacent sections versus 216.054 µm for HistoReg under comparable conditions.
Can MALDI-MSI be combined with whole-exome sequencing from a single FFPE section?
Yes. Methods have been established to extract whole-exome sequencing, RNA-seq, and metabolite-mode MALDI-MSI from a single formalin-fixed paraffin-embedded section, with greater than 91% correlation between treated and untreated transcriptomes. This enables fully integrated genomic, transcriptomic, and metabolic profiling from the same archival tissue section without requiring additional material.
Explore spatial multi-omics integration on your own research questions
BioSkepsis generates PubMed-grounded research summaries on spatial transcriptomics, MALDI-MSI workflows, proximity labeling, and biomarker validation. Every claim is citation-verified before it reaches you.
Start freeSources and further reading
- Alexandrov T, et al. SpatialData: an open and universal data framework for spatial omics. Nat Methods. 2024. PMID: 38509327
- Mund A, et al. ESCDAT enables pixel-scale co-registration of MALDI-MSI and single-cell spatial transcriptomics on the same tissue section. 2025. PMID: 41315396
- Strobl M, et al. MIIT: Multi-Omics Imaging Integration Toolset for serial-section registration of MALDI-MSI and Visium spatial transcriptomics. 2025. PMID: 40366868
- Velickovic M, et al. Spatial metabolomics of the human kidney using MALDI trapped ion mobility imaging mass spectrometry. Nat Commun. 2022. PMID: 36045222
- Branon TC, et al. Efficient proximity labeling in living cells and organisms with TurboID. Nat Biotechnol. 2018. PMID: 30125270
- Ueberheide B, et al. ProPPr: Probe-dependent Proximity Profiling identifies 1,317 phospho-tau-associated proteins in human Alzheimer's disease tissue. 2025. PMID: 40082954
- Cho KF, et al. CRISPR-based endogenous TurboID knock-in improves proximity labeling specificity at the AP1M1 locus. 2024. PMID: 39056144
- Seferbekova Z, et al. MISO: spatial multi-omics integration via spectral clustering identifies high endothelial venules in human tissue. 2024. PMID: 39815104
- Levy-Jurgenson A, et al. OmiCLIP: contrastive learning bridges histomorphology and genomics across 2 million spatial omics pairs. 2025. PMID: 40442373
- Dreisewerd K, et al. Single-cell MALDI-IHC identifies PC 36:4 enrichment in patient-derived glioblastoma tumor cells. 2025. PMID: 39932302
- Lv J, et al. Spatial multi-omic signatures predict progression-free survival in non-small cell lung cancer. 2025. PMID: 41073787
- Kaya I, et al. MALDI-MSI reveals ganglioside GM3 and GM1 co-localization with amyloid-beta plaques and GM1 d20:1/d18:1 depletion in Alzheimer's disease entorhinal cortex. Acta Neuropathol Commun. 2024. PMID: 38896306
- Milde S, et al. Correlative intravital microscopy resolves tumor microtubes and vascular invasion in glioblastoma in vivo. 2024. PMID: 39256378