Serum Protein-Based Profiles as Novel Biomarkers for the Diagnosis of Alzheimer's Disease

As a multi-stage disorder, Alzheimer's disease (AD) is quickly becoming one of the most prevalent neurodegenerative diseases worldwide. Thus, a non-invasive, serum-based diagnostic platform is eagerly awaited. The goal of this study was to identify a serum-based biomarker panel using a predictive protein-based algorithm that is able to confidently distinguish AD patients from control subjects. One hundred and fifty-six patients with AD and the same number of gender- and age-matched control participants with standardized clinical assessments and neuroimaging measures were evaluated. Serum proteins of interest were quantified using a magnetic bead-based immunofluorescent assay, and a total of 33 analytes were examined. All of the subjects were then randomized into a training set containing 70% of the total samples and a validation set containing 30%, with each containing an equal number of AD and normal samples. Logistic regression and random forest analyses were then applied to develop a desirable algorithm for AD detection. The random forest method was found to generate a more robust predictive model than the logistic regression analysis. Furthermore, an eight-protein-based algorithm was found to be the most robust with a sensitivity of 97.7%, specificity of 88.6%, and AUC of 99%. Our study developed a novel eight-protein biomarker panel that can be used to distinguish AD and control multi-source candidates regardless of age. It is hoped that these results provide further insight into the applicability of serum-based screening methods and contribute to the development of lower-cost, less invasive methods for diagnosing AD and monitoring progression.