This chapter describes the ground-breaking development of a serum-based test to help confirm the diagnosis of schizophrenia. A multiplex panel of 51 immunoassays was developed that allowed reproducible identification of schizophrenia patients compared to controls with high sensitivity and specificity. Validation of this test consisted of developing a linear support vector machine decision rule and testing its performance using cross-validation. This resulted in readjustment of the panel and algorithm to a smaller set of 40 robust assays, along with a simple procedure for maintenance and recalibration across future measurement changes associated with different reagent lots. The resulting decision rule delivered a sensitive and specific prediction for presence of schizophrenia in subjects compared to matched controls, with a receiver operating characteristic area under the curve of 88%. Performance of the recalibrated decision rule remained constant across lot changes, ensuring consistency and accuracy.

Classification, Diagnosis, Feature selection, Schizophrenia, Serum-based test, SVM,
Department of Neuroscience

Izmailov, R, Guest, P.C, Bahn, S, & Schwarz, E. (2011). Algorithm development for diagnostic biomarker assays. doi:10.1016/B978-0-12-387718-5.00011-0