Bayesian Item Response Theory for Cancer Biomarker Discovery

Abstract Item response theory (IRT), which originated from psychometric theory, relates to latent variable models for psychological assessment . IRT's contributions have been less extensively studied in cancer research even though IRT has become prevalent in the measurement of ability and item parameters with respect to latent traits. Benefits of IRT include comprehensive analyses and reduction of measurement error, meaningful scaling of latent variables, evaluation of test and item bias, greater accuracy in the assessment of change due to therapeutic intervention, and evaluation of model and person fit. Advances in cancer biology and human genetics have resulted in the identification of driver events and critical biologic dependencies. These advances, together with the development of drugs for specific biologic targets hold promise for an era of personalized oncology treatment. Therefore, robust discovery of biomarkers associated with the biologic mechanisms that drive cancers could provide clinicians with the ability to tailor therapeutic strategies for individual patients, thus improving clinical outcomes. Incorporating biomarkers in clinical trials during drug development may also allow for determining the optimal patient population for a given drug, thus maximizing its efficacy. Thus, this chapter explores the application of different IRT models in the discovery of biomarkers and altered biologic processes in cancer. The utility and limitations of the IRT models in the discovery of biomarkers or biological mechanisms are highlighted through simulations and application to genome-scale molecular profiling studies of cancers.