, researchers reported in .
“To our knowledge, this is the first time that a multianalyte plasma biomarker panel for an Alzheimer’s disease–related phenotype has been found and independently replicated by a nontargeted mass spectrometry approach,” said, of King’s College London and the University of Gothenburg in Sweden, and his research colleagues.
Blood-based measures that predict amyloid-beta burden in preclinical Alzheimer’s disease have the potential to help investigators conduct clinical trials and aid in diagnostic management. However, this novel approach needs to be validated and translated “to a simpler automated platform suitable for wider utility,” the investigators noted. In addition, it is unclear whether their classifier can track changes in amyloid-beta or differentiate between other diseases with amyloid-beta pathology.
Advances in mass spectrometry technology have renewed interest in the analysis of plasma proteins in patients with various diseases. To assess whether proteomic discovery in plasma can help predict amyloid-beta burden in preclinical Alzheimer’s disease, Dr. Ashton and his colleagues studied 238 cognitively unimpaired individuals from the Australian Imaging, Biomarker and Lifestyle Flagship Study of Ageing () and the Kerr Anglican Retirement Village Initiative in Ageing Health ( ). The participants had undergone PET to determine their amyloid-beta status. In the AIBL cohort (n = 144), 100 participants were amyloid-beta negative, and 44 were amyloid-beta positive. In the KARVIAH cohort (n = 94), 59 participants were amyloid-beta negative, and 35 were amyloid-beta positive. There were significantly more APOE4 carriers in the amyloid-beta–positive groups than in the amyloid-beta–negative groups. In addition, the amyloid-beta–positive groups tended to be older.
A support vector machine analysis created classifiers predicting amyloid-beta positivity in the AIBL cohort using demographics, proteins, or both. The researchers then tested each classifier in the KARVIAH dataset to identify which model best predicted amyloid-beta positivity. The optimal model included 10 protein features (prothrombin, adhesion G protein–coupled receptor, amyloid-beta A4 protein, NGN2, DNAH10, REST, NfL, RPS6KA3, GPSM2, FHAD1) and two demographic features (APOE4 count and age).
The classifier achieved a testing area under the receiver operator characteristic curve of 0.891 in the KARVIAH cohort to predict amyloid-beta positivity in cognitively unimpaired individuals with a sensitivity of 0.78 and specificity of 0.77.
The 10 protein features “represent a diverse array of pathways,” and the highest ranked feature was the serine protease prothrombin, which is a precursor to thrombin, the authors noted. “Multiple lines of evidence support that cerebrovascular disease may play a role in AD and that amyloid-beta may be involved in thrombosis, fibrinolysis, and inflammation via its interaction with the coagulation cascade,” the researchers wrote.
Two of the biomarkers – amyloid-beta A4 protein and NfL – have been examined in prior research and had a greater effect size in a secondary analysis that included participants with mild cognitive impairment and Alzheimer’s disease. This finding confirms “their connection with the more established disease state,” Dr. Ashton and colleagues said. In the secondary analysis, the optimal classifier included one demographic factor (APOE4 count) and nine protein features, eight of which also were used in the cognitively unimpaired classifier.
The study was funded in part by the National Institute for Health Research Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London, and many authors reported additional research support from various institutions. One author is an employee of Johnson & Johnson and a named inventor on unrelated biomarker intellectual property owned by Proteome Science and King’s College London.
SOURCE: Ashton NJ et al. Sci Adv. 2019 Feb 6. doi: 10.1126/sciadv.aau7220.