Atlas of Proteomic Technologies

n = 54,265 UK Biobank (Olink) and 35,892 Icelanders (SomaScan); 1,514 Icelanders measured on both

Sun et al. 2023 report 54,219 for the same UKB-PPP cohort.

• Compared Olink Explore 3072 in UK Biobank with SomaScan v4 in Iceland; 1,514 Icelanders were measured on both.
• A cis-pQTL was found for a similar number of assays on each platform (2,101 Olink, 2,120 SomaScan), but for a larger share of Olink assays (72% vs 43%).
• Modest cross-platform correlation: median Spearman 0.33 across 1,848 proteins in the 1,514 Icelanders measured on both.

n = 787 measured on both platforms: 568 Jackson Heart Study, 219 HERITAGE

• Compared Olink Explore with SomaScan 1.3K and 5K on precision, accuracy, breadth and phenotypic associations in 568 Jackson Heart Study and 219 HERITAGE participants.
• Modest cross-platform correlation for shared targets: median Spearman 0.44 with SomaScan 1.3K, 0.35 with SomaScan 5K.
• Olink gave more phenotypic associations, including among overlapping proteins; SomaScan gave more reproducible measurements and greater breadth.

n = 78 plasma donors (40 aged, 38 young), measured on all eight platforms

• Direct comparison of eight platforms on plasma from the same 78 donors: SomaScan 11K and 7K, Olink Explore HT and 3072, NULISA, Seer Proteograph XT, Biognosys TrueDiscovery and SureQuant targeted MS.
• SomaScan achieved best precision (median CV 5.3% for 11K, 5.8% for 7K).
• Olink Explore HT median CV ~26.8%; Seer ~26.4%; Biognosys ~29.8%

n = 427 ARIC participants measured on both platforms

Disclosure: SomaLogic ran SomaScan on ARIC samples at no charge under a collaboration agreement with Johns Hopkins, and one author is a scientific advisor to SomaLogic.

• Comparison in 427 ARIC participants at visit 5 (2011 to 2013).
• SomaScan v4 vs Olink Proseek head-to-head on the same samples: 417 protein comparisons (366 unique proteins).
• Average cross-platform correlation r = 0.46 (range -0.21 to 0.97); 19% of pairs had r ≥ 0.8.

n = 102 ARIC participants measured on both platforms

Disclosure: SomaLogic provided the assays at no cost and had no control over design or analysis; one author served on its scientific advisory committee in 2022-2023.

• Precision of SomaScan 11k and Olink Explore HT in split plasma samples from 102 ARIC participants: median CV 6.8% and 35.7%.
• Cross-platform correlations for 4,443 overlapping proteins peaked near r = 0 and r = 0.8, and one-tenth of pairs had r ≥ 0.8. The median was 0.14, below the 0.33 reported for earlier platform versions.
• Olink Explore HT precision tracks detectability (r = -0.77 with percent of samples above the limit of detection); its median CV falls to 13.3% when below-LOD values are set to half the LOD.

n = 104 participants across four Olink experiments; 53 of them also on SomaScan (14 delayed-processing, 39 sampled one year apart)

• Compared stability and reproducibility of Olink (972 proteins) and SOMAscan 7K (6,596 proteins) in archived plasma.
• Both platforms showed reproducibility in archived plasma and stability after delayed processing; over 90% of assays were stable (ICC or r ≥ 0.40) in samples collected a year apart.
• With processing delayed by up to 48 hours, more Olink than SOMAscan assays stayed stable: 55% vs 44% at ICC or r ≥ 0.75, 87% vs 72% at ≥ 0.40.

n = 1,464 Genes & Health participants; 1,396 measured on all three platforms

• Compared nanoparticle-enriched MS (Seer Proteograph XT) with SomaLogic 11k and Olink Explore HT in British South Asian participants; 1,396 were measured on all three.
• Seer detected 8,079 protein groups at least once, 5,768 of them in at least 70% of samples.
• Showed the potential of MS-based proteomics for pQTL discovery in non-European ancestries: over 1,200 locus-protein associations, more than half not previously reported.

n = 151 samples from 79 healthy donors and 72 patients (assay validation)

• Attomolar-level sensitivity: limit of detection 10 aM (0.24 fg/mL) for HIV p24 in plasma.
• Dual capture-and-release mechanism suppresses assay background, improving proximity ligation sensitivity about 10,000-fold.
• 200-plex panel was more sensitive than a proximity extension assay (Olink Explore) for low-abundance biomarkers in patients with autoimmune diseases and COVID-19.

n = 7,392 cell-culture samples from PBMCs of 6 donors (assay validation; no participant cohort)

• A 191-plex inflammation panel profiled 7,392 PBMC samples, about 1.4 million measurements.
• Antibody pairs are preassembled on barcoded beads; in spike-in testing of the 191-plex panel, 5 of more than 36,000 possible noncognate interactions were detected.
• 1,536 wells per day on one cytometer in 384-well format; 7,392 samples on a 191-plex panel in under one week.

n = 114 patients by mass spectrometry; 88 of them also by Olink

• Compared HiRIEF LC-MS/MS with Olink Explore 3072 on 88 plasma samples; 1,129 proteins were measured by both.
• Median cross-platform Spearman correlation ρ=0.59 (IQR 0.33–0.75); rises to ρ=0.68 on high-quality subset
• Platforms exhibit complementary coverage: Olink stronger at low-abundance targets; MS stronger at mid-to-high abundance

n = 500 PRECISE-SG100K participants on SomaScan; 216 also on Olink, 50 also on Seer

• Pilot for planned profiling of all 100,000 PRECISE-SG100K participants: 500 participants on SomaScan 11K, 216 of them also on Olink Explore HT, 50 also on MS-Seer.
• Three-platform comparison in a multi-ethnic Asian cohort (Chinese, Indian and Malay).
• Protein levels differed between ethnic groups. SomaScan identified more ethnicity-differentiated proteins than Olink, and 191 proteins were robust across datasets.

n = 252 plasma samples on the NULISAseq CNS panel; 218 also measured by Simoa p-tau217

• NULISAseq p-tau217 classified amyloid and tau PET status with areas under the curve of 0.92 and 0.86.
• NULISAseq p-tau217 correlated with Simoa p-tau217 (r = 0.84, n = 218), and the two assays were comparable for amyloid PET, both with areas under the curve above 0.90.
• The panel also measures ApoE4 protein, which was raised in amyloid-positive participants; the authors attribute this to the higher prevalence of APOE ε4.

n = 176 plasma samples from 113 MYHAT-NI participants on the NULISAseq CNS panel; classical biomarkers also measured by Simoa

• NULISAseq CNS panel measured 116 plasma biomarkers, with 97.2 ± 13.9% of targets above assay limits of detection.
• p-tau217 was the top marker of amyloid PET pathology (AUC 0.930 after adjustment for age, sex and APOE genotype).
• Classical Alzheimer's biomarkers correlated significantly with Simoa assays.

n = 1,260 discovery participants (Tarkin) and 325 replication samples (QMDiab) on the Seer Proteograph

• Genome-wide association study of MS proteomics using the Seer Proteograph workflow on a timsTOF Pro 2, reporting 35 previously unreported replicated pQTLs.
• Over 30% of previously reported affinity pQTLs evaluated were confirmed by MS as consistent with changes in protein abundance.
• Another 30% could not be replicated and are possibly due to epitope effects, though the authors say other explanations need case-by-case consideration.

n = 1,930 MESA participants measured on both SomaScan 7k and Olink Explore 3072

• 2,157 proteins measured on both platforms; inter-platform correlations were bimodal (median r = 0.30).
• About 25 to 30% of genetic signals on each platform are likely driven by protein-altering variants.
• 80 proteins correlated differently across ancestry groups, likely in part because protein-altering variant frequencies differ by ancestry.