Alternative normalization and analysis pipeline to address systematic bias in NanoString GeoMx Digital Spatial Profiling data
Levi van Hijfte, Marjolein Geurts, Wies Vallentgoed, Paul H.C. Eilers, Peter A.E. Sillevis Smitt, Reno Debets et al. · 2022 · 41 citationsRead the paper
Spatial transcriptomics is a novel technique that provides RNA-expression data with tissue-contextual annotations. Quality assessments of such techniques using end-user generated data are often lacking. Here, we evaluated data from the NanoString GeoMx Digital Spatial Profiling (DSP) platform and standard processing pipelines. We queried 72 ROIs from 12 glioma samples, performed replicate experiments of eight samples for validation, and evaluated five external datasets. The data consistently showed vastly different signal intensities between samples and experimental conditions that resulted in biased analysis. We evaluated the performance of alternative normalization strategies and show that quantile normalization can adequately address the technical issues related to the differences in data distributions. Compared to bulk RNA sequencing, NanoString DSP data show a limited dynamic range which underestimates differences between conditions. Weighted gene co-expression network analysis allowed extraction of gene signatures associated with tissue phenotypes from ROI annotations. Nanostring GeoMx DSP data therefore require alternative normalization methods and analysis pipelines.
What the full text shows
Working prototype- Quantile normalization outperformed other methods (modified CPM, DESeq2, gamma fit) in correcting technical bias and restoring biological signal. (NanoString GeoMx DSP data from glioma samples)
- Standard Q3 normalization fails to account for signal-to-noise ratio differences, leading to biased differential gene expression results. (Comparison between primary and recurrent glioma tumors)
- Recurrent glioma samples exhibited a systematically higher signal-to-noise ratio compared to primary samples. — p = 2.17x10-3 (Wilcoxon rank-sum test) (NanoString GeoMx DSP data)
- Biological replicates showed high reproducibility in signal distribution. — Pearson correlation > 0.98 (Eight paired samples from four patients)
- WGCNA using quantile-normalized data identified a T cell gene module highly correlated with actual T cell counts. — Pearson’s R = 0.64, p = 4x10-9 (Quantile-normalized GeoMx DSP data)
- NanoString DSP data has a limited dynamic range compared to bulk RNA-seq, underestimating differences between conditions. (Comparison of median values between bulk RNA-seq and GeoMx DSP)
- Setup
- 72 ROIs from 12 glioma samples (6 patients), 8 replicate samples, and 5 external datasets; used the NanoString Cancer Transcriptome Atlas (CTA) panel (~1800 genes).
- How
- The researchers benchmarked four normalization methods (modified CPM, DESeq2, gamma fit, and quantile) against technical and biological criteria, using bulk RNA-seq as a reference. They applied Weighted Gene Co-expression Network Analysis (WGCNA) to extract gene signatures and validated them against immunofluorescent T cell counts.
- Limitations
- Quantile normalization may limit the detection of subtle gene expression differences. · The CTA panel size (~1800 genes) may influence results. · Methodological differences between bulk RNA-seq and GeoMx DSP complicate direct comparison.
Summarized by Gemma 4 from the open-access full text · 13 Oct 2026. Read it