Abstract
Objective: Microarray-based signatures for clinical application are often plagued by processing variability or batch effects that compromise the robustness of the test performance.
Methods: A splice variant array-based signature for early detection of Alzheimer’s disease (AD) was developed using 315 AD or normal subjects processed in three disparate microarray batches.
Results: A modified top scoring pair classifier using the signature, is robust to batch effects and outperforms other common classifiers, with sensitivity and specificity of 88.3% (95% CI:81.2%, 93.4%) and 88.9% (95% CI:65.3%, 98.6%), respectively, on an independent cohort.
Conclusions: This splice-variant array-based signature shows promise for clinical diagnostic use in AD.
Acknowledgements
We would like to thank Dr Brandon W. Higgs for his statistical contributions in this study and Dr Eric Martens for scientific writing assistance.