What is Dr. Zubair Khalid's research focus?

Dr. Zubair Khalid specializes in molecular virology, mRNA vaccine development, and computational biology, with a focus on avian pathogens like IBDV and Avian Reovirus.

Where is Dr. Zubair Khalid currently working?

Dr. Zubair Khalid is a Postdoctoral Research Associate at the University of Maryland (UMD), specifically within the Department of Animal and Avian Sciences.

Spatial Single Cell RNA Seq

Spatial single cell RNA seq is a set of technologies that measure gene expression across many individual cells while preserving their physical location within a tissue section. Unlike standard single cell RNA seq, which loses spatial context during tissue dissociation, these methods let you ask where a given cell type or gene signature is located and how neighboring cells communicate. This guide is for bench scientists, bioinformaticians, and principal investigators who want to understand the core concepts, select an appropriate platform, and avoid common pitfalls when designing or analyzing a spatial transcriptomics experiment. EMBL-EBI Training provides foundational courses on bulk and single cell transcriptomics that can help you prepare, while Bioconductor offers open source packages specifically built for spatial data analysis.

Spatial single cell RNA seq sits at the intersection of histology, imaging, and high throughput sequencing. The key idea is that you capture both a transcriptomic profile and a coordinate for each detected spot or cell. The exact spatial resolution varies widely across platforms. Some methods produce spots containing 1 to 10 cells, others claim true single cell resolution. Your choice depends on the biological question, the tissue type, the gene detection sensitivity you need, and your budget. NCBI Bookshelf contains chapters on the fundamentals of transcriptomics and spatial methods that are worth reviewing before you start.

At a Glance

Aspect	Key Points
Definition	Gene expression measurement with spatial coordinates at single cell or near single cell resolution
Core technologies	Imaging based (MERFISH, seqFISH, Xenium) and sequencing based (10x Visium, Slide seq, DBiT seq)
Data output	A count matrix with spatial barcodes coordinates, often accompanied by a histology image
Main uses	Tumor microenvironment mapping, developmental biology, neuroscience, tissue atlas construction
Key challenges	Low gene detection per spot, data sparsity, computational integration of image and transcriptome
Resolution trade off	Higher resolution usually means fewer genes detected per cell or higher cost

Decision Criteria for Choosing a Platform

Before you start, clarify your biological objective. Are you looking for broad tissue zones or fine grained cell neighborhoods? The answer determines which spatial resolution and gene throughput you need.

Resolution. If you need to distinguish individual cells, choose a platform that claims single cell resolution, such as MERFISH or Xenium. If a 50 to 100 micron spot containing multiple cells is acceptable, 10x Visium is a widely used and well supported option. Galaxy Training Network has tutorials on preprocessing Visium data that can help you evaluate this approach.

Gene detection sensitivity. Imaging based methods typically query a pre selected gene panel (often 100 to 500 genes). Sequencing based methods capture polyadenylated RNA and can detect thousands of genes per spot, but with higher dropout. For discovery applications where you do not know which genes matter, an untargeted sequencing approach is usually better. For hypothesis driven work on a defined gene list, imaging may give you higher detection efficiency.

Tissue type and section quality. Fresh frozen tissue is compatible with most sequencing based methods. Formalin fixed paraffin embedded (FFPE) sections work with some platforms like Visium FFPE and Xenium. If your tissue is delicate or archived, check compatibility.

Budget and throughput. Imaging platforms require specialized equipment and often have higher per sample cost but lower sequencing cost. Sequencing based methods use standard sequencers and can scale to many samples, but the library prep cost per spot is significant. NCBI Sequence Read Archive contains public spatial RNA seq datasets that you can examine to estimate expected data sizes and quality.

Existing bioinformatics infrastructure. If your lab is already using R and Bioconductor, choose a platform with mature R packages. If you prefer Python, tools like Squidpy and Scanpy support spatial data. Bioconductor hosts dozens of packages for spatial analysis, including SpatialExperiment, BayesSpace, and SPOTlight.

Practical Workflow and Implementation Steps

The following sequence represents a typical pipeline for sequencing based spatial transcriptomics. Adjust steps according to your platform.

1. Tissue preparation and sectioning

Cryosection fresh frozen tissue at 10 to 20 micron thickness. Place the section on a slide with capture areas that contain spatial barcodes. For Visium, the capture areas contain thousands of spots, each with a unique barcode and a poly(dT) capture probe. For FFPE tissue, use a dedicated protocol with probe hybridization.

2. Staining and imaging

Stain the section with hematoxylin and eosin or immunofluorescence. Acquire a high resolution brightfield or fluorescence image. This image will be aligned to the transcriptomic data later. Store the slide carefully to avoid RNA degradation.

3. Permeabilization and cDNA synthesis

Permeabilize the tissue to release mRNA, which diffuses to the capture probes on the slide. Reverse transcription incorporates the spatial barcode into each cDNA molecule. The permeabilization time is critical too short leads to low capture, too long causes signal spreading and loss of spatial resolution. Run a permeabilization test if your platform allows it.

4. Library preparation and sequencing

Synthesize the second strand, amplify the library, and sequence on a NextSeq or NovaSeq. Aim for at least 100 million to 200 million reads per Visium slide. Paired end reads capture the spatial barcode and the transcript sequence. NCBI Sequence Read Archive provides examples of deposited spatial libraries that you can use to benchmark your own data.

5. Preprocessing and spatial alignment

Use the platform specific software (Space Ranger for Visium, or custom pipelines for Slide seq) to demultiplex reads, align them to the reference transcriptome, and generate a spot by gene count matrix. Also align the histology image to the spot coordinates. Inspect the alignment by overlaying spots on the image using tools like Loupe Browser or Bioconductor’s SpatialExperiment package.

6. Quality control

Filter spots with low total counts (e.g., fewer than 500 UMIs) or high mitochondrial content ( > 20 percent). Check for spatial artifacts such as tissue tears, bubbles, or uneven permeabilization. Visualize the number of detected genes per spot on the tissue image. Galaxy Training Network has a spatial transcriptomics quality control tutorial that walks through these checks.

7. Normalization and clustering

Normalize the count data using methods designed for spatial data, such as SCTransform or scran, accounting for tissue dependent library sizes. Perform dimensionality reduction (PCA, UMAP) and cluster spots. Because each spot may contain multiple cells, a single cluster often represents a tissue region rather than a pure cell type.

8. Cell type deconvolution

Use reference single cell RNA seq data to deconvolve each spot into estimated contributions of different cell types. Tools like SPOTlight, CARD, or RCTD are available in Bioconductor. Validation using independent markers or co detection of known cell type genes (e.g., CD3D for T cells) is essential.

9. Spatial analysis

Identify spatially variable genes (SVGs) using methods like SpatialDE, SPARK, or Seurat’s FindAllMarkers with a spatial location test. Detect tissue neighborhoods or niches by clustering spots based on both expression and spatial proximity. Compute ligand receptor interactions across neighboring spots using tools such as CellChat or NicheNet. EMBL-EBI Training covers these downstream analysis steps in its single cell and spatial transcriptomics modules.

Common Mistakes

Confusing spot resolution with single cell resolution. Most sequencing based platforms capture multiple cells per spot. Interpreting a spot as a single cell leads to erroneous conclusions about co expression and cell type purity. Always deconvolve or at least acknowledge the mixture.

Ignoring batch effects from histology. Differences in tissue processing, permeabilization time, or storage conditions can introduce systematic biases that appear as spatial patterns. Use batch correction tools (e.g., Harmony, Scanorama) but be cautious they may also remove true biological differences. PubMed study 42441075 systematically benchmarked single cell and spatial RNA seq methods for nanopore long reads, emphasizing the importance of rigorous normalization and batch effect handling.

Over interpreting lowly expressed genes. Dropout rates are high because each spot captures only a fraction of the mRNA present. A gene detected in only a few spots does not necessarily mean it is absent elsewhere. Use statistical models tuned for zero inflation, or if using an imaging platform, confirm with orthogonal staining.

Neglecting to validate with independent methods. Spatial transcriptomics can generate compelling maps, but false positives due to probe cross reactivity, alignment errors, or tissue autofluorescence are real. Validate key findings with RNAscope, immunohistochemistry, or in situ hybridization. Two recent studies used spatial transcriptomics combined with multi omics approaches and emphasized the need for downstream validation to confirm prognostic signatures and tumor microenvironment patterns 42443661, 42440056.

Running standard single cell pipelines without spatial awareness. Many analysis steps assume cells are independent and identically distributed, which is false for spatial data. Using global normalization without accounting for spatial autocorrelation can inflate false positives for spatially variable genes.

Limits of Interpretation

The resolution and sensitivity of spatial single cell RNA seq impose fundamental limits on what you can conclude. A spot covering 55 microns in diameter contains, on average, 3 to 10 cells in a typical mammalian tissue. You cannot definitively assign a transcript to a single cell unless the method is explicitly single cell resolved, and even then, optical crowding or overlapping cells may produce ambiguous assignments.

Gene detection sensitivity is lower than in standard single cell RNA seq because capture occurs on a slide surface rather than in a droplet or well. Many lowly expressed genes will be missed. This dropout biases differential expression analyses and makes it difficult to identify rare cell populations. The problem is especially acute for sequencing based platforms. Imaging based methods generally have higher sensitivity for the genes included in the panel but cannot discover new transcripts.

The computational integration of spatial transcriptomics with histology images is still an active area of research. Registration errors between the image and the barcoded spots can shift coordinates by tens of microns. Small or highly folded tissues can be especially challenging. Manually adjusting the alignment in a visual tool is common but introduces user bias.

Interpretation of spatial cell cell communication must be treated as inference, not proof. Ligand receptor pair expression in adjacent spots suggests possible signaling, but it does not demonstrate actual binding or downstream activation. Combining spatial data with perturbation experiments or functional assays is necessary to confirm. PubMed study 42443752 demonstrated the power of combining single cell and spatial transcriptomic analysis to reveal distinct tumor microenvironment signatures, but the authors noted that the spatial relationships were correlative and required further functional validation.

Finally, the field is moving quickly. New technologies and analytical methods appear every few months. A platform that is state of the art today may be obsolete next year. Keep an eye on benchmarking studies and community agnostic analysis pipelines, such as those hosted on Galaxy Training Network and Bioconductor.

Frequently Asked Questions

1. Can I use spatial single cell RNA seq to discover new cell types?

Spatial transcriptomics identifies where known cell types are located, but it is not the primary tool for discovering new cell types. Deconvolution relies on a reference single cell RNA seq dataset. If you have no reference, you can try unsupervised clustering of spots, but the resulting clusters usually correspond to tissue regions, not pure cell types. For true discovery, profiled isolated cells with single cell RNA seq first, then map them spatially.

2. How many genes are detected per spot in a typical Visium experiment?

A typical 10x Visium experiment on fresh frozen tissue detects 500 to 2,000 unique genes per spot, with median UMIs around 1,000 to 5,000. The exact numbers depend on tissue type, RNA quality, and sequencing depth. Imaging based methods like MERFISH typically detect 100 to 500 genes per cell but with higher capture efficiency.

3. What is the difference between spot resolution and pixel resolution?

Spot resolution refers to the discrete capture area on the slide (e.g., a 55 micron diameter spot for Visium). Pixel resolution is the image resolution of the histology slide, which can be sub micron. These two resolutions differ by orders of magnitude. When you overlay gene expression on the image, each spot covers many pixels. Do not mistake a pixel level detail for spot level data.

4. Do I need a computational background to analyze spatial transcriptomics data?

Most platforms provide a graphical user interface for initial quality control and visualization (e.g., Loupe Browser for Visium). However, downstream analysis including deconvolution, spatial variable gene detection, and cell cell interaction analysis typically requires programming in R or Python. If your team lacks bioinformatics support, consider partnering with a core facility or attending training courses such as those from EMBL-EBI Training.

Zubair Khalid