Untargeted Metabolomics: Plasma Panel Flags Adrenal Cancer
Telling an adrenocortical carcinoma from a benign adenoma before surgery is still guesswork more often than clinicians would like; imaging cutoffs and hormone panels leave many adrenal masses indeterminate. Plasma untargeted metabolomics offers another kind of evidence. Writing in Frontiers in Pharmacology, Yang and colleagues profiled 153 samples and built a six-metabolite model that flagged carcinoma at an AUC of 0.9266.
Key Takeaways
A six-metabolite plasma panel separated adrenocortical carcinoma from benign adenoma and healthy controls at an AUC of 0.93 in a 153-person cohort.
Glycocholic acid was the strongest single marker (AUC 0.89), placing bile acid handling at the center of the signal.
Untargeted metabolomics is a discovery tool: it ranks candidates from an open profile, then hands them to a targeted assay.
No independent cohort has tested the panel, so these are promising discovery data, not a validated clinical test.
Untargeted Metabolomics Across Carcinoma, Adenoma, and Control Plasma
Three arms were compared: 42 adrenocortical carcinomas, 56 adenomas, 55 healthy controls. EDTA plasma went through acetonitrile precipitation and 10 kDa ultrafiltration, then onto a Waters HSS C18 column with detection on a SCIEX TripleTOF 5600 and a QTRAP 6500+ in negative mode. Of 488 identified metabolites, 123 differed at VIP above 1 and FDR below 0.1. The adenoma arm carries the weight here, since that is the comparison a surgeon actually faces.
Key Findings
A six-marker model reached AUC 0.9266: 3-oxododecanoic acid, 5-hydroxy-L-tryptophan, 12-hydroxydodecanoic acid, glycine-conjugated deoxycholic acid, D-maltose, and glycocholic acid, with ten-fold cross-validation at 0.912 plus or minus 0.086.
Glycocholic acid topped the single-analyte ranking at an AUC of 0.8857, ahead of 12-hydroxydodecanoic acid at 0.8742 and trans-2-hexenyl-2-methylbutyrate at 0.8688.
Medium-chain fatty acid intermediates recurred on the shortlist, with 3-oxododecanoic acid at 0.8392, hinting at altered beta-oxidation alongside the bile acid shift.
Staging stayed out of reach: glycocholic acid did not track ENSAT pathological stage, so this reads as a detection panel, not a prognostic one.
Figure 1. OPLS-DA of the plasma metabolome across the three study groups. Panel A is the score plot, with adrenocortical carcinoma (n = 42), adenoma (n = 56), and healthy control (n = 55) samples in three largely separate confidence ellipses. Panel B is the permutation test, where the Q2 intercept falls below zero at -0.25, the pattern expected when a model is not fitting noise. Adapted from Yang et al. (2026), Frontiers in Pharmacology.
Bile Acids Carry Most of the Diagnostic Weight
Two of the six panel members are conjugated bile acids, which is biologically plausible: adrenal steroidogenesis and hepatic bile acid conjugation draw on the same cholesterol pool. There is a practical worry, though. Glycocholic acid shifts with fasting state, gallbladder emptying, and mild cholestasis, so a marker this sensitive to preanalytical handling needs standardized collection before it travels. Set against the urine steroid work of Arlt and colleagues (2011), this is a confirmatory extension more than a new concept.
What the Cohort Can and Cannot Support
Forty-two cancers is a thin discovery set for a six-variable model, and the authors say so. Cross-validation guards against the worst overfitting, yet it reuses the same 153 samples; only an independent cohort on a different LC gradient will show whether 0.93 survives. The retrospective design also blocked any correlation with Ki-67 index or tumor size, which is where clinical interest usually sits. A similar constraint shaped the plasma metabolite panel for early gastric cancer.
What This Means for Biomarker Programs
For diagnostic teams, the operational lesson is that a six-analyte signature discovered on a QTOF still has to transfer cleanly to a triple quadrupole assay, and to a second collection site, before anyone calls it a test. We hit that handoff constantly in client work at Dalton, which is why our untargeted metabolomics workflows feed a shortlist straight into quantitative confirmation on the same plasma aliquot. Programs that plan the transition up front, as disciplined multi-omics biomarker discovery efforts do, don't usually lose their best candidates at validation.
Frequently Asked Questions
What is untargeted metabolomics?
Untargeted metabolomics measures as many small molecules as the instrument can detect, without deciding in advance which ones matter. Here that meant 488 identified plasma metabolites by LC-MS. The output is a ranked candidate list, not a validated concentration.
Can untargeted metabolomics distinguish adrenocortical carcinoma from a benign adenoma?
In this cohort it did, with a six-metabolite model at an AUC of 0.9266 and glycocholic acid alone at 0.8857. The figures come from a single-center retrospective set of 42 carcinomas, so external validation is needed before such a panel could inform surgery.
How much plasma does a metabolomics biomarker study need?
Most LC-MS metabolomics workflows run on 50 to 200 microliters per injection, so archived biobank aliquots usually suffice. Consistency matters more than volume: tube type, fasting state, and time to freezing drive more variance than instrument choice. Batch correction across plates and pooled quality-control injections are standard practice.
Conclusion
Adrenocortical carcinoma leaves a detectable metabolic mark in plasma, and bile acids carry much of it. What isn't proven is that this six-marker panel holds accuracy outside the discovery set. The sensible next move is a prospective multi-center cohort with the six candidates measured by targeted metabolomics rather than rediscovered.
Related Reading
See how we run these analyses in one lab: Dalton's multi-omics CRO services.
Citation
Yang, Y., Shi, S., Wu, J., Xu, T., Shang, J., Yu, J., Zhang, B., & Liu, X. (2026). Plasma untargeted metabolomics reveals promising diagnostic metabolites for adrenocortical carcinoma. Frontiers in Pharmacology, 17, 1817514. https://doi.org/10.3389/fphar.2026.1817514
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.