Matrix matters: head-to-head concordance of serum and plasma for NULISAseq CNS Disease Panel

Blood-based proteomic profiling is now widely applied in neurodegenerative and neuroinflammatory disease, yet the choice between serum and plasma remains poorly characterised for high-multiplex platforms. Many legacy biobanks hold mainly serum, whereas most current NUcleic-acid-Linked Immuno-Sandwich Assay (NULISA) studies use plasma. We compared the 130-protein NULISAseq central nervous system (CNS) Disease Panel head-to-head in matched serum and plasma collected at the same draw from 62 participants (30 neurodegenerative, 19 demyelinating, 13 healthy controls). Agreement was measured with Spearman correlation (rho), Lin's concordance correlation coefficient (CCC), the intraclass correlation coefficient (ICC) and the mean paired serum-to-plasma difference (dNPQ). Concordance was moderate to high: 123 of 130 proteins reached significance and 18 reached rho >= 0.90, with a median rho of 0.72 (range 0.10-0.988). Proteins fell into three tiers. Cytoskeletal markers (NEFH rho=0.988; NEFL rho=0.947) and glial GFAP (rho=0.949, |dNPQ|<0.5) were interchangeable between matrices. Phosphorylated tau (pTau) species retained excellent rank concordance but carried a systematic plasma-greater-than-serum offset (pTau-181 rho=0.869, dNPQ=+0.67; pTau-217 rho=0.846, dNPQ=+0.64; pTau-231 rho=0.885, dNPQ=+0.89). Platelet-derived analytes (CD40LG rho=0.102, dNPQ=-4.74; BDNF rho=0.223, dNPQ=-2.69) and intracellular synaptic proteins (NRGN, SNAP25, ENO2) diverged markedly. For most clinically relevant neurodegeneration markers, especially cytoskeletal and glial proteins, serum is a valid substitute for plasma; absolute thresholds for phosphorylated tau and amyloid peptides require matrix-specific calibration, and platelet-sensitive analytes cannot be compared across matrices without strictly standardised pre-analytical conditions.