Abstract Details

Title Reanalysis of Global Neurodegeneration Proteomics Consortium Data Uncovers Pervasive Risk of False Positives and False Negatives in Biomarker Discovery from Large, Multi-center Proteomic Studies

Topic Movement Disorders

Presentation(s) P6 - Poster Session 6 (5:00 PM-6:00 PM)

Poster/Presentation
Number 16-014

Objective

To develop high-plex proteomic data analysis methods for fluid biomarker discovery in the setting of neurodegeneration.

Background The Global Neurodegeneration Proteomics Consortium (GNPC) V1 harmonized dataset harbors measurements for 7,289 unique proteins from multiple Somalogic assays across >35,000 biofluid samples from Alzheimer’s disease, frontotemporal dementia, Amyotrophic Lateral Sclerosis, and Parkinson’s disease (PD) patients, plus healthy controls (HCs). Published case-control analyses per cohort and disease (Imam 2025) yielded median chi-squared test statistics (proteomic inflation factors) ranging from 1.01 to 26.29 (median: 1.47), suggesting statistical miscalibration.

Design/Methods We reanalyzed GNPC blood cohorts with >20 PD cases and >80 HCs. PD participants with coincident diseases were excluded. Protein markers below the limit of detection in >12.5% of samples or correlated with cell type composition were excluded. Protein abundances were standardized relative to HCs accounting for sex, age, and average protein abundance. We performed logistic regressions on case status against protein abundance, sex, years of education, and the first 5 principal component scores. Cohorts with hits were meta-analyzed via weighted Z-test.

Results The number of Bonferroni significant meta-analyzed markers decreased from 2,251 (Imam 2025) to 28. Proteomic inflation factors all fell below 1.93 (median: 1.13). The rate of significant biomarker sign discordance across different cohorts decreased from 38% to 1%. The ordering of candidate biomarkers ranked by significance changed dramatically (rho = 0.04); for example, integrin aVb5 (a heterodimer of frequently reported PD biomarkers ITGAV and ITGB5), did not possess even nominal significance before but ranked 16th in our meta-analysis.

Conclusions

Choice of statistical approach profoundly impacts qualitative and quantitative results of biomarker association to neurodegeneration case-control statuses. Proteomic inflation factors may flag possible confounding in large, multicenter studies. Protein abundance standardization, removal of cases with coincident disease, and incorporating principal components as covariates can assist rigorous protein biomarker discovery in large-scale, multi-cohort proteomic datasets.

Authors/Disclosures
Evan A. Boyle, PhD PRESENTER	Mr. Boyle has received personal compensation for serving as an employee of Octave Bioscience. Mr. Boyle has received personal compensation for serving as an employee of Exai Bio. The institution of Mr. Boyle has received research support from Michael J Fox Foundation.
Katrina Paumier, PhD (Mitsubishi Tanabe Pharma America (MTPA))	Dr. Paumier has nothing to disclose.
Ludmila Voloboueva, PhD	Dr. Voloboueva has received personal compensation for serving as an employee of Octave Bioscience.
Ferhan Qureshi (Octave Bioscience)	Mr. Qureshi has received personal compensation for serving as an employee of Octave Bioscience.
David Brazel, PhD	Dr. Brazel has received personal compensation for serving as an employee of Octave Bio. Dr. Brazel has or had stock in Octave Bio. The institution of Dr. Brazel has received research support from MJFF. The institution of an immediate family member of Dr. Brazel has received research support from DOE/NASA/DOD.

��ɫ��

��ɫ��