Beyond Cell Typing: Protein-Guided Cell Annotation and Spatial Phenotype Discovery with SpatialX

The Next Bottleneck in Spatial Biology Isn’t Data Generation

Spatial omics technologies such as Visium, Xenium, CosMx, and multiplexed imaging platforms such as PhenoCycler have transformed how researchers study tissues by enabling molecular measurements within their native spatial context.

These technologies now routinely generate:

• Gene × cell expression matrices
• Protein × cell intensity matrices
• Spatially resolved tissue maps

For most research groups, the challenge is no longer data generation.

It is an interpretation.

Researchers can now profile thousands to millions of cells within intact tissues, but extracting biologically meaningful insight remains difficult. One of the most time-consuming and variable steps in spatial analysis is still cell type annotation—the conversion of high-dimensional molecular measurements into interpretable cellular identities.

Figure 1: Spatial omics technologies such as Visium, Xenium, CosMx, and multiplexed imaging platforms such as PhenoCycler have transformed how researchers study tissues by enabling molecular measurements within their native spatial context 

Cell Type Annotation Remains the Critical Bottleneck
Before spatial relationships can be studied, raw expression data must be transformed into cell-level annotations.

However, in current workflows, this step is often:

• Dependent on manual marker selection
• Sensitive to threshold choices
• Inconsistent across datasets and studies
• Difficult to scale across large cohorts

Whether working with transcriptomic or proteomic data, researchers face the same fundamental challenge: “how to reliably map expression signals to biologically meaningful cell types and states”.

This step is essential but often fragmented across tools and subjective in execution.

SpatialX: Protein-Guided Cell Type Annotation for Spatial Omics

SpatialX introduces a new approach to one of the most important steps in spatial analysis: cell type annotation driven by protein expression, with integrated support for transcriptomic context.

Rather than relying solely on manual gating or unsupervised clustering, SpatialX models expression structure directly from the data to guide annotation of both cell identities and functional states.

This enables researchers to move from raw data:

• Gene × cell matrices (spatial transcriptomics)
• Protein × cell matrices (imaging proteomics)

to:

• Consistent, biologically meaningful cell annotations
• Functional immune and stromal states
• Curated, reproducible phenotype definitions

A human-in-the-loop annotation system

Cell annotation requires both computational structure and biological expertise. SpatialX combines automated modeling with interactive refinement:

• Interactive curation and adjustment of markers

• Automated detection of expression structure

• Suggested protein-driven cell type boundaries

• Probabilistic assignment of cell identities

Researchers remain in control while avoiding the need to reconstruct annotation strategies from scratch for each dataset.

This reduces variability while preserving biological interpretability.

Figure 2: Human-in-the-loop protein-guided annotation workflow in SpatialX

Why Protein-Guided Annotation Matters

Accurate annotation depends on selecting molecular features that reflect true cellular identity and function.

Protein expression is often more directly linked to cellular phenotype than RNA alone, particularly in immune and stromal compartments where functional states are reflected at the protein level.

Many canonical cell markers—including surface receptors, lineage markers, and activation markers—are measured at the protein level.

By combining both modalities, SpatialX enables a more complete view of cellular identity:

• Protein signals define functional execution and surface cell states
• RNA signals reveal underlying transcriptional and regulatory programs

Integrated analysis allows researchers to better resolve cases where molecular layers agree—or diverge ¹. This is especially important in complex tissues where cellular identity alone is not sufficient to explain functional behavior.

RNA–Protein Decoupling as a Source of Biological Insight

A growing body of spatial multi-omics studies shows that RNA and protein expression are not always aligned ^1–3.

Recent work by Weiss et al. (2026) ¹ demonstrates that RNA–protein concordance varies systematically depending on gene function, pathway context, and cellular state. Their findings suggest that:

• Proteomic organization reflects functional specialization across cell types
• RNA–protein divergence follows pathway-specific patterns 
• Transcriptomic and proteomic measures of cell identity can diverge in biologically meaningful ways

These observations reinforce an important principle: “Biological identity cannot be fully captured by RNA or protein measurements alone.” 

As spatial technologies increasingly enable simultaneous measurement of multiple molecular layers, understanding the relationships between RNA, protein, and spatial context becomes essential for accurately characterizing cellular function and tissue-level biology.

In many tissues, RNA–protein uncoupling can reveal:

• Post-transcriptional regulation
• Protein turnover dynamics
• Immune activation states not reflected in RNA
• Spatially restricted protein expression programs
• Functional heterogeneity within transcriptionally similar cells

SpatialX enables researchers to compare protein-guided annotations with transcriptomic cell-state predictions in the same spatial context.

Figure 3: Evaluate CD8A protein expression for cell type annotation

Figure 4: Cell type prediction based on transcriptional layer

Comparing these complementary views allows researchers to identify where molecular layers agree—and where they diverge—providing additional insight into cellular regulation and tissue organization.

From Annotated Cells to Spatial Phenotypes
Once protein-guided and transcriptomic annotations are established, researchers can begin exploring tissue architecture at a systems level.

A spatial phenotype is a biological state defined not only by cellular identity, but also by functional activity and spatial context. Rather than describing only what a cell is, spatial phenotypes capture what cells are doing, where they are doing it, and how they interact with neighboring cellular communities.

SpatialX supports analyses that connect molecular interpretation with spatial organization, including:

• Cellular neighborhoods defined by functional states
• Immune infiltration and exclusion patterns
• Tumor–stroma spatial interfaces
• Spatial gradients of activation and suppression
• Multi-layer tissue organization across RNA and protein

These analyses help transform annotated cells into biologically meaningful spatial phenotypes.

From Data to Discovery

Spatial omics is transforming biology from a cell-centric view of tissues toward a systems-level understanding of tissue organization. As datasets become increasingly multi-modal, the challenge is no longer simply identifying what cell types are sitting on a slide. The larger objective is understanding how molecular programs, functional states, and spatial relationships interact to shape biological behavior.

Protein-guided annotation provides an important foundation for this transition by connecting molecular measurements to biologically grounded cellular states. When integrated with transcriptomic information and spatial coordinates, these annotations enable researchers to move beyond descriptive cell maps toward the discovery of functional tissue architectures.

Ultimately, the value of spatial omics lies not only in measuring more molecules, but in understanding how molecular states are spatially organized to generate tissue-level function. As RNA, protein, and spatial measurements become increasingly integrated, the ability to interpret these relationships will define how effectively complex biological systems can be understood.

Explore Your Spatial Omics Data with SpatialX

Ready to see how multi-layer spatial analysis can accelerate discovery in your datasets? Access the SpatialX platform to test our available public reference datasets or upload your own cohort stacks to start your discovery pipeline.

Request a Managed SpatialX Demo or Start Your Code-Free Trial today.

References

1. Weiss, C. A. M., Sjöstedt, E., Debnath, A., et al. “A cell type-resolved proteomic atlas of the human body,” bioRxiv, 2026, p. 2026.05.26.727663.

2. Samih, A., Ferreira, M. A. de M., and Nikoloski, Z. “Gene expression and protein abundance: Just how associated are these molecular traits?,” Biotechnology Advances, V. 86, 2026, p. 108720.

3. Sandoval-Filarsky, C., Glock, C., López, J. M. A., et al. “Soma-seq links intracellular protein states to transcriptional programs in the human brain,” BioRxiv, 2026.