Correlation between antibody- and aptamer-based plasma proteome measurements

Olink Explore HT and SomaScan 11k are two widely used affinity-based platforms for large-scale plasma proteomics, differing in their underlying detection chemistries, assay designs, and target coverage. To assess the concordance between measurements from both platforms, Spearman rank correlations were calculated for all proteins targeted by common assays across both technologies, using 2,621 matched plasma samples from the same individuals. In SomaScan, individual proteins can be targeted by multiple aptamers with distinct epitope specificity, resulting in multiple protein-aptamer pairs available for some target proteins.

The distribution of per-protein-aptamer Spearman correlations is bimodal, with a large fraction of proteins showing near-zero correlation and a distinct second peak at approximately 0.8, indicating that a subset of proteins is highly concordant between the two platforms while another subset shows limited agreement.

This pattern is consistent with observations in previous cross-platform comparisons and likely reflects differences in epitope specificity, assay sensitivity, and protein isoform recognition between the two technologies.

Figure 1. Histogram showing the distribution of Spearman correlations across protein-aptamer pairs.

Disease-specific correlation


Many proteins measured by both platforms are annotated as intracellular and are typically present at low levels in circulation under baseline conditions. This can lead to low overall correlations, which may improve in disease contexts where protein levels are higher. To explore this, correlations were also calculated within individual disease groups. This revealed that, for a subset of proteins, cross-platform agreement increases substantially in specific disease contexts. When using the highest disease-specific correlation per protein, the overall distribution shifts toward higher agreement, with nearly half of protein-aptamer pairs showing strong correlation. This pattern is consistent with increased detectability in disease samples and suggests that platform concordance depends on the biological context and signal strength.


Figure 2. Heatmap comparing Spearman correlations of protein-aptamer pairs between baseline and disease-specific conditions.


Table 1. Spearman correlations per protein-aptamer pair, calculated across all 2,621 samples (Baseline) and within individual disease groups (highest value shown).

Gene Gene description Olink SomaScan Spearman
Baseline
Spearman
Disease specific
Diff
CRKL CRK like proto-oncogene, adaptor protein OID44571 9877-28 0.96 0.97 0.01
TMED8 Transmembrane p24 trafficking protein family member 8 OID44328 30328-1 0.96 0.97 0.01
TWF2 Twinfilin actin binding protein 2 OID45291 23396-21 0.96 0.97 0.01
NCK2 NCK adaptor protein 2 OID44804 13615-60 0.95 0.97 0.02
MKNK1 MAPK interacting serine/threonine kinase 1 OID43977 23693-4 0.95 0.97 0.02
GP6 Glycoprotein VI platelet OID45136 3194-36 0.95 0.97 0.02
ERP29 Endoplasmic reticulum protein 29 OID44636 13728-19 0.95 0.96 0.01
PRTFDC1 Phosphoribosyl transferase domain containing 1 OID42956 20950-159 0.95 0.96 0.01
TBCB Tubulin folding cofactor B OID44310 17686-27 0.95 0.96 0.01
DTD1 D-aminoacyl-tRNA deacylase 1 OID44613 19194-9 0.95 0.96 0.01
CHMP1A Charged multivesicular body protein 1A OID43527 18224-11 0.95 0.96 0.01
MANF Mesencephalic astrocyte derived neurotrophic factor OID44766 9223-11 0.95 0.96 0.01
MGMT O-6-methylguanine-DNA methyltransferase OID43971 21437-77 0.95 0.97 0.02
PYM1 PYM homolog 1, exon junction complex associated factor OID44885 29989-145 0.95 0.96 0.01
CNP 2',3'-cyclic nucleotide 3' phosphodiesterase OID43546 6609-22 0.95 0.96 0.01
BDNF Brain derived neurotrophic factor OID45040 14047-78 0.95 0.97 0.02
Show allShow less


Protein detection reliability as a driver of cross-platform correlation

A key determinant of cross-platform concordance is the proportion of samples in which a protein is reliably detected above the limit of detection (LOD). For each protein, a sample was classified as below LOD if either the Olink or the SomaScan measurement fell below the respective platform's LOD threshold; the proportion of Olink measurements below LOD was substantially greater than for SomaScan. The distribution of proteins by percentage of samples above LOD is strongly bimodal, with the majority of proteins either detected in fewer than 10% of samples or detected reliably across all samples. Median Spearman correlation increases with the fraction of samples above LOD, reaching values above 0.5 for proteins detected in more than 90% of samples, demonstrating that detection reliability is a primary driver of cross-platform agreement. Proteins with low detection rates in either platform should therefore be interpreted with caution when comparing measurements across technologies. More information about sample selection, LOD definitions, and correlation methodology can be found in the Method summary.

Figure 3. Trends in Spearman correlations by assay detectability. The histogram shows the number of proteins at different detectability levels, defined as the percentage of samples per protein above LOD in either Olink or SomaScan. The line plot shows the median correlation at each detectability level.