Clinical Proteomics: Panel Confirms Vasculitis Remission
- Aug 9
- 5 min read
Deciding when to stop immunosuppression in ANCA-associated vasculitis (AAV) is guesswork more often than anyone treating it would like. C-reactive protein (CRP) and ANCA titer stay elevated in plenty of patients who are clinically quiet, so therapy drags on and infection risk climbs with it. A clinical proteomics study in Nature Communications went at that gap directly: plasma from 163 people was profiled by DIA LC-MS/MS, and the seven-protein panel that came out of it told active disease from remission with an AUC of 0.94 in an independent cohort recruited in another country.
Key Takeaways
A seven-protein plasma panel (AHSG, CLEC3B, COMP, F9, LRG1, MCAM, MRC1) separates active ANCA-associated vasculitis from remission more accurately than CRP or ANCA titer.
Clinical proteomics yields decision-support assays, not just protein lists, when discovery profiling is followed by a targeted assay tested in a separate cohort.
Correcting for CRP, kidney function, and blood counts before feature selection stops a candidate panel from collapsing into a proxy for generic inflammation.
At 30 minutes per sample, targeted mass spectrometry sits inside the throughput range clinical laboratories actually work at.
Building a Clinical Proteomics Panel in Four Steps
Discovery quantified 605 plasma proteins, and disease status explained 34.2% of proteome variance (PERMANOVA, p = 0.001) while freezer time, MS plate, and run order each accounted for 1.2% or less. Active versus remission comparison returned 268 differentially expressed proteins at 5% FDR. Instead of ranking those by fold change, the authors ran MetaDeconfoundR against CRP, eGFR, hemoglobin, hematocrit, platelets, and ANCA titer; only 135 held a disease association not reducible to a measured covariate. LASSO trimmed that to 21, each measured by PRM with heavy-labeled internal standards, and Boruta selection landed on seven. Plasma proteome profiling of this shape has been standard since Geyer and colleagues (2016), but the confounder-aware step is what makes this panel something other than an expensive CRP.
Key Findings
Seven proteins survived two rounds of selection: AHSG, CLEC3B, COMP, F9, LRG1, MCAM, and MRC1, with MRC1, LRG1, and F9 tracking active disease and the rest tracking remission.
Performance held across cohorts: AUC 0.98 (95% CI 0.92 to 1.00) in the discovery test partition, 0.94 (0.90 to 0.98) in an independent 108-patient validation cohort, 0.97 (0.95 to 1.00) combined.
It beat the markers clinicians already use: ANCA titer reached AUC 0.75 and CRP 0.93, and the panel gave 5 false positives against 13 for CRP (specificity 88.1% versus 69.0%).
Conventional markers mislead often: among patients in BVAS-defined remission, 65.7% remained ANCA-positive and 20.0% had elevated CRP, yet panel discrimination stayed at AUC 0.97 in exactly those patients.
Figure 1. Study design. Step 1, global plasma proteomics by LC-MS/MS in a discovery cohort of 50 healthy controls plus 60 active and 54 remission AAV patients. Step 2, confounder-aware feature selection with MetaDeconfoundR, then LASSO-penalized logistic regression. Step 3, verification by targeted PRM mass spectrometry with Boruta reduction to a seven-protein signature. Step 4, generalized linear modeling applied to an independent validation cohort. Adapted from Jerke et al. (2026), Nature Communications.
Where CRP and ANCA Titer Fall Short
Roughly three quarters of patients scored as being in remission still had at least one conventional marker elevated. That single number frames the clinical problem better than any AUC does. Decision curve analysis indicated the panel would let clinicians taper immunosuppression in more genuinely quiet patients without accepting more missed activity, and adding CRP or ANCA titer back into the model improved nothing.
What the Confounder-Aware Design Actually Buys
Stripping out variance attributable to acute-phase reactants before selecting features is the part worth copying. COMP makes the case: plasma levels fell in active disease, pointing at vascular smooth muscle injury rather than inflammation, and a fold-change ranking would likely have buried it. The caveats are real. Remission labels came from BVAS, a partly subjective 56-item score, so some patients called remission may have carried subclinical activity, and the 20% discovery test set held only 22 patients with near-complete separation, which is why the external cohort carries most of the evidentiary weight.
Implications for Biomarker Programs
The transferable lesson is that the discovery platform and the deployment platform don't have to be the same instrument, provided the handoff is designed in from day one rather than bolted on after the paper. We see the same pattern in client work at Dalton, where untargeted discovery is followed by a targeted plasma proteomics assay anchored on heavy-labeled peptide standards, since absolute quantification is what lets a panel survive the move to a second site. Teams scoping a multi-omics biomarker discovery program should budget for that verification tier at the start; panels that skip it tend to shrink when they meet an independent cohort.
Frequently Asked Questions
What is clinical proteomics?
Clinical proteomics measures proteins in patient samples such as plasma, serum, or tissue to answer a diagnostic, prognostic, or treatment-response question. Most of it runs on mass spectrometry, either untargeted discovery across hundreds of proteins or targeted assays quantifying a defined shortlist. Discovery generates the hypothesis; the targeted stage turns it into a usable test.
Can a blood test show whether ANCA vasculitis is in remission?
This study says a seven-protein plasma panel can, reaching an AUC of 0.97 across the combined cohorts with better specificity than CRP. It is not a cleared clinical test, and it was evaluated only in patients already diagnosed with AAV. Prospective work on flare prediction is underway.
How much sample does a clinical proteomics biomarker study need?
Plasma volumes are small, a few microliters per injection for both DIA discovery and targeted PRM runs. Cohort size matters far more: this work used 163 discovery participants and 108 validation patients for a rare disease. The optimized assay ran in 30 minutes per sample, which sets the practical ceiling on cost per patient.
Conclusion
Two independent cohorts and a covariate-controlled pipeline make the case that this signature reflects vasculitis activity itself, not the inflammation that accompanies it. Whether the assay reads the same way in another laboratory, or holds against disease controls outside AAV, is still open. For groups running biomarker programs in rare inflammatory disease, the design is worth borrowing even if these particular proteins prove specific to this indication.
Related Reading
See how we run these analyses in one lab: Dalton's multi-omics CRO services.
Citation
Jerke, U., Kirchner, M., Bartolomaeus, T. U. P., Kling, L., Kratky, V., Hruskova, Z., Tesar, V., Eckardt, K.-U., Schreiber, A., Forslund, S. K., Mertins, P., & Kettritz, R. (2026). Towards a proteomic plasma biomarker panel for diagnosing vasculitis remission. Nature Communications, 17, 6825. https://doi.org/10.1038/s41467-026-75755-6
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.