Multi-Omics: Plasma Panels Detect Three Cancer Types
- Aug 16
- 4 min read
Blood tests for early cancer detection keep failing at the same hurdle: a marker that looks strong against healthy donors fades once it meets patients with symptoms but no tumor. A Swedish group measured 165 plasma proteins and 244 NMR metabolite parameters in 2,066 people, hunting for multi-omics combinations of two to four analytes able to beat assays already in the clinic.
Key Takeaways
A four-protein plasma panel separated colorectal cancer patients from healthy controls at AUC 0.89, matching approved stool-based screening tests.
In this multi-omics study, NMR metabolites added nothing to detection but tracked tumor stage closely.
Ovarian cancer detection reached AUC 0.97 using CA125 with plasminogen, and 0.92 with CA125 excluded.
Case-control sampling flatters screening performance, so these panels still need prospective testing in asymptomatic people.
Multi-Omics Profiling of 2,066 Plasma Samples
Samples came from U-CAN, a Swedish cohort that banks plasma at diagnosis, before treatment. After filtering on blood frozen within four hours of draw, 818 confirmed cancers remained (330 colorectal, 304 lung, 184 ovarian) alongside 119 patients whose workup of the same organ found premalignant or benign disease. Protein profiling used a Bio-Plex suspension array of 165 antigens; the metabolite side ran on the Nightingale NMR platform, mostly lipoprotein subclass measures. The authors skipped penalized regression, searching exhaustively over every combination of one to four analytes and every cutoff.
Key Findings
Four proteins, one colorectal signature: CEACAM5, FLT1, IL19 and ferritin reached AUC 0.89 against 748 healthy controls, above Epi proColon (0.82) and FIT (0.88).
Lung detection leaned on a down-regulated marker: reduced FNDC5 with elevated midkine, PLAUR and CEACAM5 gave AUC 0.91 across 304 cases.
CA125 helped but was not indispensable: MUC16 with plasminogen reached 0.97 in ovarian cancer; swapping MUC16 for plasminogen, KLK6, midkine and CCL2 returned 0.92.
Metabolites reported stage, not presence: none reached significance for detection, while 111 analytes correlated with stage in lung cancer and stage IV separated from stages I to III at AUC 0.87 to 0.92.
Against a harder comparator the numbers dropped: versus the 119 non-malignant patients, the panels fell to 0.81, 0.89 and 0.92.
Figure 1. Study design and cohort composition. Panel A traces the workflow, from plasma taken in 937 cancer cases and non-cancer controls plus 1,129 healthy controls, through affinity proteomics and NMR metabolomics, to composite biomarkers tested against a predicate device. Panel B shows selection of cases from 3,821 U-CAN patients and controls from 23,670 EpiHealth participants. Adapted from Akerren Ogren et al. (2026), Molecular Cancer.
Why the Metabolite Layer Underperformed at Detection
Not one of the 244 metabolite parameters reached significance for separating cancer cases from non-cancer controls. The authors are candid about why: fasting status and sample handling differed between U-CAN and EpiHealth, and NMR lipoprotein measures respond to both. That reads as a statement about sampling design rather than metabolite biology, and a fasting-matched replication would settle it. Stage told a different story, with signatures mixing CEACAM5, lipoprotein phospholipid percentages and GlycA flagging stage IV disease.
Checking the Protein Panels Against CancerSEEK
External support came from the CancerSEEK dataset described by Cohen and colleagues (2018), which shares 32 proteins with this panel. Across 388 colorectal, 104 lung and 54 ovarian stage I to III cases plus 812 controls, the overlapping biomarkers held similar AUC values, a confirmatory result more than a discovery. Only part of the combination space could be checked, since this study measured four times as many proteins. Protein profiling has held up elsewhere, in work on plasma proteins that predict multiple disease risks.
Running Both Omics Layers From One Aliquot
Splitting a plasma aliquot between an affinity proteomics platform and an external NMR facility is where much of the pre-analytical noise in this design creeps in. Measuring proteins and metabolites from the same injection, which the Omni-MS platform at Dalton was built around, holds both layers to one freeze-thaw history and one batch structure. That isn't a small detail for multi-omics biomarker discovery programs, since a panel pairing CEACAM5 with a lipoprotein-derived metabolite inherits whatever handling difference separated the two measurements, and single-injection multi-omics analysis removes that confounder before modeling begins.
Frequently Asked Questions
What is multi-omics analysis used for?
Multi-omics analysis measures two or more molecular layers, such as proteins and metabolites, in one set of samples. In biomarker work it finds combinations that outperform any single marker and separates disease signal from technical variation. Here, proteins carried detection and metabolites carried stage.
Can multi-omics blood panels detect colorectal, lung and ovarian cancer?
Four-protein panels reached AUC 0.89 for colorectal and 0.91 for lung cancer, and two proteins reached 0.97 for ovarian cancer. Those values match or exceed approved stool-based colorectal tests. Because the cohorts were case-control and sampled at diagnosis, the results argue for further testing, not screening use today.
How many samples does a proteomics and metabolomics biomarker study need?
This study used 818 cancer cases, 119 disease controls and 1,129 healthy controls, a realistic scale for regulatory-grade comparisons. Pilots of 60 to 100 samples per group can rank candidates but rarely survive multiple-testing correction. Standardizing collection costs less than adding cases.
Conclusion
Small protein combinations, measured on an affinity platform, can match approved screening assays in plasma drawn at diagnosis. What isn't established is anything about screening itself, since enriched case-control sampling flatters every number here. Detection programs should budget for fasting-standardized collection at the outset, because that decision determines whether the metabolite layer is usable at all.
Related Reading
See how we run these analyses in one lab: Dalton's multi-omics CRO services.
Citation
Akerren Ogren, J., Ekstrom, J., Rameika, N., Torell, E., Larsson, C., Stoimenov, I., Micke, P., Gyllensten, U., Hellstrom, M., Glimelius, B., Stalberg, K., & Sjoblom, T. (2026). Composite proteomic and metabolomic plasma biomarkers for detection of colorectal, lung and ovarian cancers. Molecular Cancer, 25(1), 105. https://doi.org/10.1186/s12943-026-02654-1
Note
This blog post summarizes findings from the above-cited research. Figures are adapted from the original publication. For full details, please refer to the source article.
By Seungjun Yeo, CEO at Dalton Bioanalytics. Specializing in multi-omics mass spectrometry for drug discovery and biomarker research.