PUBLICATIONS - Abstract Details

Multiplex Panels as Diagnostic Tools in Alzheimer’s Disease [abstract]
Soares, Holly, Keystone Symposia on Molecular and Cellular Biology; Alzheimer’s Disease Beyond Abeta, 2010 January 10-15; Copper Mountain, Colorado (CO); Keystone Symposia; 2010. p 62. Abstract 024.

Alzheimer's disease is a progressive neurodegenerative disorder estimated to impact more than 27 million worldwide. Currently, the clinical diagnosis of AD is based upon presentation of cognitive symptoms, usually memory, and significant social and/or day-day dysfunction that can be characterized as dementia. Emerging data suggest that CSF abeta42 and tau and imaging (e.g. amyloid and FDG PET and MRI atrophy) biomarkers can differentiate AD from controls and other forms of dementia. Furthermore, these same biomarkers have shown utility in identifying symptomatic, pre-demented patients who then progress to dementia. As a result, there is considerable evidence to believe Alzheimer's disease can be identified in pre-demented stages with the aid of biomarkers. While CSF and imaging markers are amenable as research tools for clinical trial usage, these platforms pose issues as standard of care population screening tools. Effective prevention will likely need a non-invasive cost-effective tool to identify early AD. As a result, efforts have focused upon attempting to identify blood based signatures of Alzheimer's disease. The current approach utilized a 190 analyte quantitative validated immunoassay based multiplex panel to identify potential diagnostic markers of Alzheimer's disease using over 2000 plasma and CSF samples from AD, controls and other dementia. The general strategy involved training the models using one of the datasets and then testing the performance in a subset of the original dataset and in 2 additional independent sample sets using different sites. Preliminary data suggest a baseline demographic model including age, gender and ApoE provides sensitivity of 80% and specificity of 65% which is typical of sample sets where enrollment is based upon a pre-specified diagnosis of AD. Boosted tree, random forest and support vector machine approaches were then used for feature selection and plasma models from the multiple panels, when used in combination with the baseline demographic model, yielded a sensitivity and specificity of 80% respectively. Many of the analytes identified in the plasma panel were also identified in the CSF as markers that were differentially expressed in AD vs. controls. In summary, a plasma based signature for AD appears to be present and may have utility as a screening tool for monitoring subjects at risk of AD.