Variability of 7K and 11K SomaScan plasma proteomics assays
Julián CandiaiD, Giovanna FantoniiD, Francheska Delgado‐PerazaiD, Nader Shehadeh, Toshiko TanakaiD, Ruin MoaddeliD, Keenan A. WalkeriD, Luigi FerrucciiD
Abstract
SomaScan is an aptamer-based proteomics assay designed for the simultaneous measurement of thousands of human proteins with a broad range of endogenous concentrations. The 7K SomaScan assay has been recently expanded into the new 11K version. Following up on our previous assessment of the 7K assay, here we expand our work on technical replicates from donors enrolled in the Baltimore Longitudinal Study of Aging. By generating SomaScan data from a second batch of technical replicates in the 7K version, as well as additional intra- and inter-plate replicate measurements in the new 11K version using the same donor samples, this work provides useful precision benchmarks for the SomaScan user community. Beyond updating our previous technical assessment of the 7K assay with increased statistics, here we estimate inter-batch variability, we assess inter- and intra-plate variability in the new 11K assay, we compare the observed variability between the 7K and 11K assays (leveraging the use of overlapping pairs of technical replicates), and explore the potential effects of sample storage time (ranging from 2 to 30 years) in the assays precision.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 56
- FWCI
- 7.60
- Field percentile
- top 1%
- Organism
- Assay
- Research Focus
- Approach
In-depth summary
Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (25 Sept 2024) · CC0 1.0
Across Baltimore Longitudinal Study of Aging replicates, fully normalized SomaScan 11K has median inter-plate coefficient of variation 5% (7K: 4.5%), intra-plate 3%, 7K inter-batch 8%.
SomaScan, an aptamer-based plasma proteomics assay increasingly used for biomarker discovery, recently grew from 7K to 11K reagents, raising the need to re-benchmark its precision, batch sensitivity and robustness to long biobank storage. Extending the authors' earlier 7K duplicate study, this work adds a second 7K batch plus 11K intra- and inter-plate replicates from the same donors, enabling a paired cross-version comparison.
Key results
ShownThe 11K assay keeps the precision of the 7K version. The standard normalization pipeline lowers technical variability further with each step it adds.
Fully normalized median inter-plate CV 5% for 11K (82 duplicates) vs 4.5% for 7K, with per-SOMAmer CVs strongly correlated across 7,289 shared SOMAmers; 11K median CV fell from 12% (raw RFU) to 11% (hybridization), 8% (plate-scale), 7% (inter-plate calibration) and 5% (full normalization). · Results and Discussion; Conclusions; Figure legends
ShownPlate and batch layout shape precision. Replicates on the same plate agree more closely than replicates on different plates. Replicates from different batches are clearly noisier, even after both batches were renormalized together.
11K median CV 3% intra-plate vs 5% inter-plate, correlated across 10,776 SOMAmers but shifted off the y = x line; 7K median inter-batch CV 8% vs 4.5% intra-batch inter-plate, although both batches (2021, 2022) were renormalized jointly with the same calibrators. · Results and Discussion; Conclusions; Figure legends
SuggestedFor most SOMAmers, storage time (2 to 30 years) had no significant link to replicate variability. The few targets that were associated were protein-specific and positively correlated, which fits moderate degradation.
7K: 128 inter-plate pairs stored 2 to 28 years; only seven proteins passed FDR q = 0.05, all positive (top: COL6A1, TYRP1); aging markers PTN and GDF15 showed mild or no effect. 11K: 82 pairs stored 5 to 30 years, none significant. 11K storage-affected (690, nominal p < 0.05) and freeze-thaw-affected (466) SOMAmers overlapped by 31 (Fisher p = 0.77). · Results and Discussion; Conclusions; Figure legends
ShownAgreement within replicate pairs across SOMAmers is close to perfect in both versions. However, comparing the same pairs run on both versions shows slightly higher correlation in 7K than in 11K.
74 inter-plate pairs run on both versions; mean Spearman correlation difference (11K minus 7K) Δr = −0.0051, Wilcoxon paired p = 0.0013; a small significant negative difference remained when 11K was limited to the 7,289 shared SOMAmers and after dropping five outlier samples. · Results and Discussion; Conclusions; Figure legends
ShownPrecision differs by dilution group and only partly tracks signal-to-background ratio (SBR). Almost all proteins sit well above buffer background. CV and SBR showed no association with protein properties such as length, mass or number of interactions.
11K CV vs SBR Spearman r = 0.48 overall; 0.51 (20% dilution, n = 8644), 0.13 (0.5%, n = 1911), −0.17 (0.005%, n = 221). Median CV 5 ± 1.5%, 6 ± 2%, 7 ± 2% for the 20%, 0.5%, 0.005% groups (pairwise Wilcoxon adjusted p < 0.05). 98% of proteins >2-fold brighter than buffer wells. · Results and Discussion; Conclusions; Figure legends
Methods, data and reproducibility
BLSA plasma was profiled with SomaScan 7K (v4.1; two batches) and 11K (v5.0) using intra- and inter-plate technical duplicates; per-SOMAmer coefficients of variation (CVs) were estimated by grid search for the CV whose theoretical log-normal fold-change distribution best fits the empirical replicate distribution under a one-sample Kolmogorov-Smirnov test.
- BLSA EDTA plasma (stored at −80 °C) run as inter-plate technical duplicates on SomaScan 7K (v4.1) in two batches (2021 and 2022) · human · n = 128 inter-plate pairs (storage analysis) · 10.17605/OSF.IO/2B9QE
- Same BLSA donor samples rerun on SomaScan 11K (v5.0) in 2024 as intra- and inter-plate duplicates · human · n = 82 inter-plate duplicates (82 subjects) · 10.17605/OSF.IO/2B9QE
- Inter-plate replicate pairs measured on both 7K and 11K for paired cross-version correlation analysis · human · n = 74 inter-plate pairs
- UniProtKB annotations of 11K SOMAmer targets (length, mass, interactions, subcellular location, other properties) · human · n = 10,893 SOMAmer-UniProt pairs; 9,609 UniProt IDs
- Methods & tools
- SomaScan 7K (v4.1), SomaScan 11K (v5.0), SomaScan normalization (hybridization, median signal, plate-scale, inter-plate calibration), Grid-search CV estimation with one-sample Kolmogorov-Smirnov test, Spearman correlation and Wilcoxon paired test, Fisher's exact test, R v.4.4.0, queryup v.1.0.5, UniProtKB, RColorBrewer, viridis, calibrate, BioRender
- Code & protocols
- https://osf.io/2b9qe
- Data & accessions
- https://osf.io/2b9qe
- Software & environment
- R v.4.4.0; queryup v.1.0.5; RColorBrewer v.1.1-3; viridis v.0.6.5; calibrate v.1.7.7
- Reproducibility gaps
- Anonymized data and custom R code are shared on OSF, but no container, lockfile or compute requirements are described beyond the listed R package versions.
Limitations (5)
- Many samples were not aliquoted for single-use thawing, and historical records are incomplete. Repeated freeze-thaw exposure therefore cannot be tracked and may be mixed up with storage-time effects. · stated by the authors
- Calibrator-based normalization did not remove cross-batch effects. The authors flag batch correction (bridge samples, ComBat, guided PCA, multi-MA normalization) as needing further investigation. · stated by the authors
- Storage-linked degradation appears protein-specific and is not significantly associated with SOMAmers affected by freeze-thaw (31 shared, Fisher p = 0.77), so the authors note that a correction procedure would be complex to devise. · stated by the authors
- All replicates come from a single cohort's EDTA plasma processed under one laboratory's protocols, so the precision benchmarks may not carry over to serum, other collection protocols or other sites. · noted in this summary
- The per-SOMAmer storage analyses correlate replicate fold change with storage time without covariate adjustment, so storage time may be confounded with donor age, collection era or pre-processing changes, none of which are modeled. · noted in this summary
Who should read:Read this if you design or analyze SomaScan plasma studies and need plate, batch, version (7K vs 11K) and storage-time precision benchmarks to plan sample layouts or interpret longitudinal biobank data.
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Assessment of Variability in the Plasma 7k SomaScan Proteomics Assay
Julián Candia, Gulzar N. Daya, … Keenan A. Walker·Scientific ReportsPeer-reviewed·Oct 13, 2022
- SomaScan Bioinformatics: Normalization, Quality Control, and Assessment of Pre-Analytical Variation
Julián Candia·Methods in molecular biologyPeer-reviewed·Jan 1, 2025
- Effects of in vitro hemolysis and repeated freeze-thaw cycles in protein abundance quantification using the SomaScan and Olink assays
Julián Candia, Giovanna Fantoni, … Luigi Ferrucci·Journal of Proteome ResearchPeer-reviewed·Apr 18, 2025
- Systematic Development Of Sandwich Immunoassays For The Plasma Secretome
R. S. Haeussler, A. Bendes, … J. M. Schwenk·bioRxiv·Jan 8, 2019
- High-throughput plasma proteomic platforms: Insights from a multi-ethnic Asian cohort
Akash Bahai, He Wang, … Nilanjana Sadhu·bioRxiv·Oct 24, 2025
- Application of Cancer Cell Line Encyclopedia for Measuring Correlation Between Transcriptomics and Proteomics as a Guide for System-level Insights
Blake Williams, Darryl Perry, … Joseph Loureiro·bioRxiv·Mar 3, 2024