- Spatial bioinformatics has transformed biological research by enabling the investigation of gene expression, proteins, metabolites, and cellular organization within their original tissue environments. Technologies such as spatial transcriptomics, imaging-based transcriptomics, spatial proteomics, and spatial multi-omics provide information that conventional bulk molecular analysis cannot capture. However, these technologies have important limitations related to spatial resolution, molecular sensitivity, measurement accuracy, data interpretation, and computational complexity. Understanding these limitations is essential for designing reliable experiments, selecting appropriate analytical methods, and avoiding incorrect biological conclusions.
- Spatial resolution refers to the level of detail at which molecular information can be assigned to a particular location, cell, or subcellular structure. Different spatial technologies provide different levels of resolution. Some platforms measure gene expression across relatively large tissue spots containing multiple cells, while others identify individual cells or individual RNA molecules. Spot-based technologies such as conventional 10x Visium and some sequencing-based spatial methods may combine signals from several neighboring cells. Consequently, it can be difficult to determine which cell type produced a particular transcript or whether a detected expression pattern represents one cell population or a mixture of several populations. This problem is especially important in complex tissues such as tumors, brain, liver, and inflamed organs, where different cell types may occur close together.
- Higher-resolution platforms attempt to overcome this limitation by measuring molecular information at smaller spatial scales. Technologies such as Visium HD, Xenium, MERFISH, seqFISH, and CosMx can provide finer spatial measurements, with some supporting cellular or subcellular localization. However, higher nominal resolution does not automatically guarantee accurate identification of every cell or transcript. Actual performance depends on probe design, transcript detection efficiency, image quality, tissue preservation, segmentation accuracy, and the molecular coverage of the platform. In addition, the physical size of a cell, the density of the tissue, and the distribution of RNA within the cell can affect how reliably molecular measurements are assigned to individual cellular structures.
- Cell segmentation is another important source of spatial-resolution uncertainty. Cell segmentation is the computational process of identifying the boundaries of individual cells in microscopy images. If segmentation is inaccurate, transcripts or proteins may be assigned to the wrong cell, neighboring cells may be incorrectly merged, or a single cell may be divided into multiple artificial objects. These errors can affect cell-type identification, cellular neighborhood analysis, and cell–cell communication predictions. Tools such as Cellpose, StarDist, CellProfiler, and QuPath support segmentation, but their performance depends on image quality, tissue type, staining, and the biological structures being analyzed. Manual inspection and validation are often necessary, particularly in densely packed or morphologically complex tissues.
- Spatial sensitivity refers to the ability of a technology to detect molecules that are actually present in a particular tissue location or cell. Not every RNA molecule, protein, or metabolite is detected during an experiment. Low-abundance transcripts may remain undetected, while weak protein signals may fall below the detection threshold. This can produce incomplete molecular profiles and make it difficult to identify rare cell populations, lowly expressed genes, or subtle differences between tissue regions. Detection sensitivity depends on the molecular abundance, probe or antibody performance, sequencing depth, imaging conditions, sample preparation, and the technical characteristics of the platform.
- Transcript capture efficiency is a particularly important limitation in spatial transcriptomics. Some RNA molecules may be degraded during tissue processing, fail to bind to capture probes, or remain undetected during sequencing or imaging. As a result, measured gene expression may represent only a fraction of the molecules present in the tissue. This can create sparse expression matrices containing many zero values. A zero measurement may indicate genuine absence of expression, but it may also result from failure to detect a transcript that was present. This distinction is important when identifying marker genes, estimating cell types, detecting rare transcripts, and performing differential expression analysis.
- Probe-based imaging technologies have their own sensitivity limitations. Their performance depends on probe specificity, hybridization efficiency, background fluorescence, signal amplification, and the number of genes included in the assay. Some platforms measure predefined gene panels rather than the entire transcriptome, meaning that genes outside the selected panel cannot be detected. This can limit the discovery of unexpected biological pathways or rare cell states. Probe cross-hybridization and nonspecific binding can also generate misleading signals, while high background fluorescence can reduce the ability to distinguish genuine molecular detections from noise.
- Spatial proteomics faces additional sensitivity and specificity challenges. Antibody-based approaches depend on antibody quality, epitope accessibility, staining conditions, and signal detection. Antibodies may show cross-reactivity or inconsistent performance between experiments. Multiplexed protein imaging also involves technical constraints related to the number of proteins that can be measured simultaneously, signal overlap, tissue autofluorescence, and repeated staining or imaging procedures. Mass spectrometry imaging can detect a broad range of molecular species, but metabolite and lipid identification may be difficult because of overlapping mass signals, ion suppression, matrix effects, and differences in ionization efficiency.
- Tissue preparation can strongly influence spatial measurements. Fixation, freezing, sectioning, permeabilization, staining, and storage may affect the preservation and accessibility of RNA, proteins, and metabolites. RNA degradation can reduce transcript detection, while tissue distortion can alter spatial coordinates and anatomical structures. Some molecular components may be lost during processing, and different protocols may introduce systematic differences between samples. These effects are particularly important when comparing disease and control tissues, different experimental batches, or samples collected from different laboratories.
- A major challenge in spatial bioinformatics is interpreting molecular measurements from complex tissue environments. A spatial expression pattern may reflect changes in cell-type composition rather than changes in gene expression within individual cells. For example, an increase in an immune-associated gene within a tumor region may result from the accumulation of immune cells rather than increased expression by tumor cells. This is known as a cellular composition effect. Single-cell reference data and computational deconvolution methods can help distinguish these possibilities, but the accuracy of the estimates depends on the quality and representativeness of the reference dataset.
- Spatial deconvolution estimates the proportions or abundance of different cell types within a mixed spatial measurement. Methods such as RCTD, cell2location, SPOTlight, stereoscope, and DestVI can help resolve cellular mixtures in spot-based spatial transcriptomics. However, their results are model-dependent and may be affected by incomplete reference profiles, similar gene-expression patterns between cell types, platform differences, and variation in cell states. A reference dataset generated from a different tissue condition or species may not accurately represent the cells present in the spatial sample. Deconvolution results should therefore be interpreted as estimates rather than exact measurements of cellular composition.
- Spatial proximity also creates challenges in interpreting cell–cell communication. Cells located near one another may not necessarily communicate directly, and the detection of ligand and receptor transcripts does not establish that the corresponding proteins are present, active, or interacting. Signaling may depend on protein processing, receptor localization, post-translational modifications, intracellular pathways, and other molecular factors. Some communication mechanisms operate over longer distances, while others require direct cell contact. Computational tools such as CellChat, CellPhoneDB, NicheNet, and LIANA identify candidate interactions, but experimental validation is necessary to confirm important biological mechanisms.
- Another limitation concerns the distinction between correlation and causation. Spatially associated genes, proteins, metabolites, or cell types may occur together because they respond to the same environmental condition rather than because one directly influences another. For example, two cell populations may be enriched in the same tissue region because both respond to inflammation, without communicating directly. Similarly, a gene-expression pattern associated with tumor progression may be a consequence rather than a cause of disease development. Spatial associations provide valuable biological hypotheses, but causal relationships generally require functional experiments, perturbation studies, or additional mechanistic evidence.
- Statistical interpretation is also challenging because spatial measurements are not always independent. Neighboring cells and tissue locations may share biological and technical characteristics, producing spatial autocorrelation. Standard statistical methods that assume independent observations may underestimate uncertainty or produce misleading significance values if spatial dependence is ignored. Multiple testing is another concern because spatial studies may examine thousands of genes, cell types, regions, and molecular interactions. Appropriate statistical models, permutation testing, false discovery rate correction, and biological replication are important for reliable conclusions.
- Batch effects and platform differences can further complicate interpretation. Measurements generated using different instruments, tissue-processing protocols, sequencing platforms, imaging systems, or laboratories may show technical variation unrelated to biology. Integration methods such as Seurat integration, Harmony, fastMNN, and scVI can help reduce unwanted technical variation in suitable datasets. However, excessive correction may remove genuine biological differences, particularly when comparing distinct tissue regions, disease states, or cell populations. Researchers should therefore evaluate whether integrated datasets preserve known biological signals and meaningful spatial patterns.
- Biological variability is another important consideration. Different individuals, animals, tissue regions, developmental stages, and disease conditions may exhibit genuine differences in cellular composition and molecular activity. A small number of tissue samples may not capture this variability. Furthermore, a single tissue section may not represent the entire organ or tumor because biological processes can vary substantially across different regions. Spatial studies should therefore use appropriate biological replicates and sampling strategies that reflect the complexity of the research question.
- Spatial multi-omics introduces additional challenges because different molecular layers may have different resolutions, sensitivities, and coordinate systems. Transcriptomic, proteomic, genomic, epigenomic, and metabolomic measurements may be collected using separate technologies or from adjacent tissue sections. Integrating these datasets requires normalization, image registration, feature matching, and consideration of differences in molecular coverage. Small registration errors may create false spatial associations, while missing measurements in one molecular layer can limit joint analysis. Computational frameworks such as Seurat, MOFA+, scvi-tools, Squidpy, Giotto, and SpatialExperiment support different aspects of multi-omics and spatial integration, but no single framework resolves every technical challenge.
- Computational complexity and data management are also significant limitations. High-resolution spatial experiments can generate large imaging files, sequencing datasets, molecular matrices, segmentation masks, and spatial graphs. Processing these data requires suitable computing resources, storage, memory, and reproducible analytical workflows. Deep-learning models may require substantial training data and computational capacity, while differences in software versions and parameter settings can affect reproducibility. Standardized data formats, metadata, quality-control procedures, and workflow documentation are important for sharing and comparing spatial datasets.
- Several strategies can improve the reliability of spatial bioinformatics studies. Researchers should select technologies according to the biological question and the required molecular and spatial resolution. Sample preparation and tissue handling should be standardized, and quality-control metrics should be assessed throughout the experimental workflow. Cell segmentation and molecular detection should be checked against tissue images and established markers. Suitable single-cell reference datasets can improve cell-type mapping and deconvolution, while comparisons across multiple analytical methods can reveal whether important results are robust. Statistical analyses should account for spatial dependence, multiple testing, biological replication, and technical variation.
- Independent experimental validation is essential for important findings. RNA in situ hybridization, RNAscope, immunofluorescence, immunohistochemistry, multiplexed protein imaging, and spatial proteomics can validate molecular localization and cellular identity. Functional experiments, including gene knockdown, gene editing, receptor inhibition, and cell-culture or tissue-model studies, can help establish whether predicted molecular interactions have biological effects. Validation should be designed around the specific claim being tested, since confirmation of gene expression does not necessarily establish protein activity or causal function.
- Overall, limitations in resolution, sensitivity, and interpretation remain important challenges in spatial bioinformatics. Spatial technologies provide unprecedented information about molecular organization within tissues, but their measurements are affected by detection efficiency, tissue preparation, cell segmentation, platform characteristics, statistical assumptions, and biological complexity. Understanding these limitations helps researchers distinguish reliable observations from uncertain predictions and design stronger analytical and experimental workflows. Continued improvements in spatial resolution, molecular detection, computational modeling, standardization, and experimental validation will increase the reliability and biological value of spatial omics research.