Exploratory proteomic biomarker combinations associated with post-concussion symptoms following a concussion in adolescents
Frontiers in Neurology, 2026
Gill J., Lim A., Dennis E., Bickart K., Yun S., Yun J., Alice J., Miles C., Chrisman S., Vaughan C., Cullum C., Cook L., Gioia G., Rivara F., Giza C., Bazarian J.
| Disease area | Application area | Sample type | Products |
|---|---|---|---|
Neurology | Patient Stratification | Plasma |  Olink Explore HT |
Abstract
Objective
This study examined the ability of established brain biomarkers, glial fibrillary acid protein (GFAP), neuro-filament light chain (NFL), ubiquitin carboxy hydrolase-L1(UCH-L1), tau, and phosphorylated tau (p-tau), and novel biomarkers from blood collected sub-acutely after concussion to determine prediction of persisting post-concussion symptoms (PPCS) beyond 3 months. We hypothesized that a combination of established and novel proteins would predict high PPCS burden.
Participants
Adolescents 11 to 17.99 years with a concussion based on Concussion In Sport Group (CISG) criteria were eligible. Participants were assessed 7–35 days post-injury (baseline) and then reassessed at 3 months (follow-up) for persistent post-concussion symptoms (PPCS) using the Post-Concussion Symptom Inventory, 2nd Edition (PCSI-2). A total of 155 participants (78 females, 77 males) with both blood biomarkers and 3-month symptom data were analyzed.
Design
Plasma proteins collected sub-acutely (7–35 days post concussion) were quantified by both Quanterix and the Olink Explore platform, and compared between participants with the highest and lowest quartiles of PPCS severity at follow-up (85–95 days post concussion) using the PCSI-2. An exhaustive best-subsets logistic regression strategy was executed following clinical and biological pre-filtering to identify parsimonious multivariable configurations of protein biomarkers distinguishing individuals with high and low PPCS at follow-up. A stratified 10-fold cross-validation framework was implemented to evaluate model generalizability and safeguard against overfitting, while bootstrapping was used to calculate confidence intervals. Ingenuity Pathway Analysis (IPA) was performed to generate hypotheses of molecular pathways implicated in PPCS pathogenesis.
Results
None of the brain biomarkers collected sub-acutley were significantly different in PPCS-high and PPCS-low groups at 90-day follow-up. No Olink proteins survived multiple testing correction, but a cross-validated multivariable model, including Tripartite Motif Containing 39 (TRIM39) + Sclerostin (SOST) + Transcription factor Dp family member 3 (TFDP3) + TNF receptor superfamily member 9 (TNFRSF9) + Leptin (LEP) + Prune Homolog 2 With BCH Domain (PRUNE2) + Trimethylguanosine Synthase 1 (TGS1) + EPCAM (Epithelial Cell Adhesion Molecule) distinguished high PPCS at follow-up with an area under the curve (AUC) of 0.86 (95% CI 0.80–0.92). IPA identified three significant networks associated with PPCS: (a) cardiovascular and neurological disease, organismal injury, and abnormalities (score = 45; 28 focus molecules); (b) connective tissue disorders, inflammatory disease, and organismal injury and abnormalities (score = 41; 26 focus molecules); and (c) connective tissue development and function, embryonic development, and organismal development (score = 41; 26 focus molecules).
Conclusion
This study suggests the possibility to utilize novel biomarkers discovered by high throughput proteomic analysis to predict high PPCS. Future research should further develop precision of unique biomarker profiles and prolonged symptomatology in adolescents’ post-concussion.