Accessible Sampling for Diagnostic Proteomics

Evidence

Two recent studies tested whether proteins collected through relatively accessible samples could sharpen difficult diagnostic distinctions. They examined different diseases and used different technologies, so they do not validate one shared biomarker. Together, however, they show that sample collection is only the first step: preparation, tissue compartment, comparator groups, and validation design determine what a proteomic signal can mean.

The first study compared five preparation workflows for liquid chromatography–tandem mass spectrometry of human plasma. Using plasma from five healthy controls, the researchers tested SP3, iST, and ENRICH-iST on raw plasma; SP3 after depletion of 14 highly abundant proteins; and SP3 after enrichment of plasma extracellular vesicles. Raw-plasma SP3 quantified a mean of 391 proteins. Depletion followed by SP3 quantified 646, while extracellular-vesicle enrichment followed by SP3 quantified 923. The two deeper workflows also captured different protein subsets: 152 proteins were exclusive to depleted plasma and 161 to the extracellular-vesicle workflow.

The team then used those two workflows on plasma from 15 women with relapsing-remitting multiple sclerosis at diagnosis, before treatment, and five sex- and age-matched healthy controls. Depleted plasma yielded 54 regulated proteins and the extracellular-vesicle analysis yielded 35. Only four overlapped: A2M, IGKV3-15, APOA2, and HYOU1. Von Willebrand factor was elevated and proposed as a candidate diagnostic marker, but the study did not establish a diagnostic test. The low overlap indicates that changing plasma preparation changed both measurement depth and the biological signal available for interpretation.

The second study used tape strips rather than blood to sample inflammatory skin disease. This prospective observational study included 14 adults with hyperkeratotic palmoplantar eczema, 10 with palmoplantar psoriasis, and 12 with palmoplantar pustulosis. Investigators collected lesional and non-lesional skin and measured a targeted panel of 1,034 proteins with Olink Reveal, with selected validation by multiplex immunoassays. The three conditions can look similar clinically, making them a stringent test of molecular discrimination.

Protein patterns partly separated palmoplantar pustulosis from the other two diagnoses, but psoriasis and eczema overlapped. Lesional skin across all three groups showed higher inflammatory-protein expression than non-lesional skin. Pustulosis had a distinct profile enriched for neutrophil degranulation, innate immune activation, and Th17-associated signaling centered on IL-17A and CXCL8. By contrast, no proteins were significantly different between psoriasis and eczema lesions; those groups shared interferon-associated chemokines CXCL10 and CXCL11. In a longitudinal psoriasis subset, inflammatory protein levels declined with clinical improvement. Because the source was available to this review at abstract depth, the size of that subset and detailed effect estimates are not inferred here.

The studies therefore produced mixed, informative results. Plasma fractionation expanded coverage and exposed largely nonoverlapping candidate lists, while tape strips captured a local inflammatory signal that distinguished one disease pair better than another. Neither result supports using the reported proteins as a clinical diagnostic assay.

Analysis — Accessibility Does Not Ensure Specificity

The cross-study inference is that accessible sampling and diagnostic specificity are separate engineering problems. This is analysis, not a conclusion directly tested across both diseases. Plasma is convenient compared with cerebrospinal fluid, but abundant circulating proteins can mask lower-abundance signals; depletion and extracellular-vesicle enrichment effectively create different views of the same sample. Tape strips bring measurement closer to diseased skin, yet local inflammation was not automatically diagnosis-specific: psoriasis and eczema still shared a substantial interferon-associated pattern. The studies converge on the importance of preserving compartment information, but they do so in complementary ways. The plasma study separated soluble and vesicle-enriched fractions before untargeted mass spectrometry, whereas the skin study compared lesional with non-lesional sites using a targeted panel. A useful emerging design would treat collection site and preparation workflow as part of the biomarker definition, then lock the assay, features, and decision threshold before external testing. That proposal remains unproven. Stronger evidence would require independent cohorts containing realistic disease mimics, prespecified performance measures, and demonstrations that the signal remains stable across laboratories and patient subgroups.

Limitations

The plasma study was small and single-center, with 15 RRMS cases and five healthy controls; all participants used for disease profiling were women. Its discovery comparisons relied on unadjusted significance, and healthy controls do not test specificity against inflammatory or neurological conditions that resemble multiple sclerosis. The same five controls also contributed to workflow testing. Greater protein coverage is not itself proof of better diagnostic performance, and von Willebrand factor may reflect processes that are not unique to RRMS.

The tape-strip study included 36 participants across three diagnoses. Its abstract reports partial separation but not a locked classifier, external validation cohort, or diagnostic sensitivity and specificity. This review cannot assess unreported effect sizes, correction procedures, treatment effects, or longitudinal timing from abstract-level evidence. The failure to separate psoriasis from eczema may reflect shared biology, limited power, assay choice, or all three.

Finally, the studies differ in disease, tissue, platform, and endpoint. They do not replicate one another, and their protein lists should not be combined. Their shared contribution is methodological: accessible samples can support proteomic discovery, but clinical discrimination still depends on how the sample is partitioned, what comparison is posed, and whether the resulting signature survives prospective validation.