RNA-Protein Concordance in Spatial Profiling: Why Multi-Omics Matters

We often treat RNA expression as a proxy for protein. In spatial biology, that assumption breaks down — fast.

A gene is transcribed into RNA. That RNA is translated into protein. It sounds straightforward, but in practice, the relationship between RNA and protein levels is anything but simple. Think of RNA as the blueprint and protein as the final product — but with delays, edits, and sometimes missing pieces along the way.

In spatial biology, this disconnect has real consequences. Researchers using platforms like NanoString’s GeoMx DSP (Merritt et al., 2020) can now measure both RNA (via Whole Transcriptome Atlas) and protein (via Immune Proteomics Atlas) from the same tissue sections. When they do, an uncomfortable truth often emerges: RNA and protein levels don’t always agree.

The Concordance Problem

When we say “RNA-protein concordance,” we’re asking a deceptively simple question: if a gene’s RNA is highly expressed, is its protein also highly expressed?

The answer varies dramatically depending on the gene, the tissue compartment, and the biological context.

Genome-wide RNA-protein correlations are typically modest — first demonstrated in yeast (Gygi et al., 1999) and later confirmed in mammalian systems (Schwanhausser et al., 2011), with median Pearson r values around 0.4-0.6. In our spatial profiling work, we’ve seen concordance drop to around 60% for some biomarkers in certain datasets — meaning that 40% of the time, RNA and protein tell different stories. For some immune checkpoint molecules, the concordance is strong and reliable. For others, particularly markers that are subject to post-transcriptional regulation, protein turnover, or translational repression, the relationship breaks down.

Why the Disconnect?

Several biological mechanisms drive RNA-protein discordance:

• Post-transcriptional regulation: MicroRNAs and RNA-binding proteins can suppress translation without affecting mRNA levels (Bartel, 2004)
• Protein stability: Some proteins are rapidly degraded, while others persist long after their mRNA has been cleared — protein half-lives range from minutes to weeks (Eden et al., 2011)
• Secretion and shedding: Membrane-bound proteins can be cleaved and released, reducing detectable protein at the cell surface while RNA levels remain high
• Compartment effects: A protein produced in one tissue compartment may diffuse or be transported to another, creating spatial mismatches between RNA and protein signals

What This Means for Biomarker Research

The implications for biomarker development are significant. Consider a scenario where you’re evaluating a therapeutic target — say, a surface receptor that an antibody-drug conjugate is designed to bind. If you rely solely on RNA expression to stratify patients, you might misclassify a substantial fraction:

• RNA-High / Protein-Low: These patients would be predicted to respond but may not, because the drug target isn’t actually present at the protein level
• RNA-Low / Protein-High: These patients might be excluded from treatment despite having the target, because stable protein persists even after transcription decreases

A Critical Asymmetry in RNA-Protein Relationships

A pattern consistently reported in the literature is asymmetric discordance: high RNA often predicts high protein, but low RNA does not reliably predict low protein (Liu et al., 2016). This asymmetry has significant implications for patient stratification — it means false negatives (missing patients who actually have the target) are a bigger risk than false positives when using RNA alone.

Practical Strategies for Multi-Omics Integration

1. Run Both Assays When Possible

If your experimental design and budget allow, measure both RNA and protein from the same tissue sections. The GeoMx platform makes this feasible through sequential profiling. Having both modalities lets you identify which markers are concordant (and therefore reliable as single-assay biomarkers) versus discordant (and therefore requiring multi-omics validation).

2. Assess Concordance Per Compartment

Don’t assume that concordance observed in one tissue compartment applies to others. In our experience, a marker might show strong RNA-protein correlation in tumor epithelium but weak or even negative correlation in the immune compartment. Always assess concordance within each compartment separately.

3. Use Categorical Agreement, Not Just Correlation

Pearson or Spearman correlations between RNA and protein are useful but can be misleading. A more clinically relevant metric is categorical concordance: if you classify samples as High vs. Low based on RNA, how often does the protein classification agree? This directly addresses the question of whether RNA can serve as a surrogate for protein in patient stratification.

4. Consider Composite Biomarker Scores

When single markers show poor RNA-protein concordance, consider building composite scores that integrate multiple markers. A score combining RNA expression, protein expression, and cell type composition is more robust than any individual measurement.

The Case for Multi-Omics as Standard Practice

The era of single-analyte biomarker studies is ending. As spatial profiling platforms continue to expand their multi-omics capabilities — adding epigenomics, metabolomics, and single-cell resolution — the ability to integrate across data types will become a core competency for translational research teams.

These challenges aren’t just theoretical — they show up consistently in real spatial datasets. At Cytogence, we’ve built our GeoMx analysis pipelines to handle multi-omics integration from the ground up. Our workflows include automated concordance assessment, compartment-specific analysis, and composite scoring — because understanding where RNA and protein agree (and where they don’t) is essential for turning spatial data into actionable biology.

References

1. Gygi SP, Rochon Y, Franza BR, Aebersold R. Correlation between protein and mRNA abundance in yeast. Molecular and Cellular Biology. 1999;19(3):1720-1730. doi: 10.1128/MCB.19.3.1720. PMID: 10022859.

2. Schwanhausser B, Busse D, Li N, et al. Global quantification of mammalian gene expression control. Nature. 2011;473(7347):337-342. doi: 10.1038/nature10098. PMID: 21593866.

3. Liu Y, Beyer A, Aebersold R. On the dependency of cellular protein levels on mRNA abundance. Cell. 2016;165(3):535-550. doi: 10.1016/j.cell.2016.03.014. PMID: 27104977.

4. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell. 2004;116(2):281-297. doi: 10.1016/S0092-8674(04)00045-5. PMID: 14744438.

5. Eden E, Geva-Zatorsky N, Issaeva I, et al. Proteome half-life dynamics in living human cells. Science. 2011;331(6018):764-768. doi: 10.1126/science.1199784. PMID: 21233346.

6. Merritt CR, Ong GT, Church SE, et al. Multiplex digital spatial profiling of proteins and RNA in fixed tissue. Nature Biotechnology. 2020;38(5):586-599. doi: 10.1038/s41587-020-0472-9. PMID: 32393914.

---

Cytogence provides end-to-end bioinformatics support for spatial transcriptomics and multi-omics studies. Get in touch to discuss your project.