A mass spectrometry-based discovery and replication of a multi-analyte classifier for neocortical amyloid pathology

Nicholas J. Ashton, Alejo J. Narvardo-Holgado, Steven Lynham, Malcolm Ward, Veer Bala Gupta, Pratishtha Chatterjee, Kathryn Goozee, Eugene Hone, Steve Pedrini, Ashley I. Bush, Christopher C. Rowe, Victor L. L. Villemagne, David Ames, Colin L. Masters, Dag Aarsland, Simon Lovestone, Ralph Martins, Abdul Hye

    Research output: Contribution to journalMeeting abstractpeer-review


    Background: Neocortical amyloid burden (NAB) is an important risk factor for Alzheimer’s disease that precedes the onset of clinical symptoms. It has become critical to identify individuals at early stages of NAB for the accurate recruitment of participants to clinical trials. Blood-based biomarkers predicting NAB would have great utility as a “first-step enrichment” of participants for such trials. Only a few of studies have investigated blood-based biomarkers predicting NAB, with one study utilising Mass Spectrometry (MS). MS has several advantages as discovery tool over panel-based assays but has limitations. We recently demonstrated a methodology to increase sensitivity, dynamic range and protein coverage that utilises MS. Here, we present a discovery and replication of a plasma protein panel predicting NAB.Methods: A proteomic workflow combining immunodepletion, peptide labelling and isoelectric focusing was established as a sensitive strategy for in-depth plasma exploration. Protein identification and relative quantitation was performed by an LTQ Orbitrap Velos. This was applied to 297 participants from the AIBL and KARVIAH cohorts with Aβ PET. LASSO and combined with SVM built to predict Aβ SUVR. The model was trained and classification performance analysed in the training set (AIBL) with a 100 repeats of 5-fold cross validation. Subsequently the optimum model was tested independently in the KARVIAH cohort.Results: A refined proteomic strategy has demonstrated a reproducible workflow that can profile >1000 plasma proteins, accurately detecting low picogram levels. Preliminary analysis has demonstrated a large number of single protein markers that correlate with NAB. Furthermore, a machine learning analysis revealed a 14-analyte panel that is able to predict NAB in a cognitively normal cohort at an accuracy of 86.6%. Pathway analysis highlights the convergence of pathways involved in coagulation, APP processing, neuronal transcription factors and axonal injury to be important in predicting NAB.Conclusions: MS-based proteomics is a powerful hypothesis testing tool. We have, for the first time, performed a large LC-MS/MS discovery and replication study to accurately predict NAB. Further replication and disease understanding of these putative markers will need to be elucidated. Furthermore, the translation to an assay-based platform would be required for future clinical and research application.
    Original languageEnglish
    Article numberP3-243
    Pages (from-to)P1033
    Number of pages1
    JournalAlzheimer's and Dementia
    Issue number7 Supplement
    Publication statusPublished - 2017
    EventAlzheimer's Association International Conference 2017 - London, United Kingdom
    Duration: 15 Jul 201720 Jul 2017


    Dive into the research topics of 'A mass spectrometry-based discovery and replication of a multi-analyte classifier for neocortical amyloid pathology'. Together they form a unique fingerprint.

    Cite this