Abstract:
|
Detection of biomarkers is an important task in the context of drug discovery in both clinical and non clinical studies. A large amount of research has been devoted to the identification of surrogate endpoints and biomarkers for a primary endpoint in a setting where the biomarker is a part of a high dimensional data setting. So far, the methodology for biomarker detection was focused on detecting a single biomarker. We proposed a new modelling framework, based on a joint model for the biomarker(s) and the primary endpoint, in which multiple biomarkers can be identified from a collection of potential biomarkers. The proposed modelling framework can be used to identify a subset of K biomarkers which can be used together as a biomarker for the primary endpoint of interest (joint surrogacy) and allows to estimate the surrogacy effect of the Kth biomarker, given the surrogacy effect of (K-1) biomarkers (partial surrogacy). Orthogonal surrogacy is a special case of partial surrogacy, consisting of K independent biomarkers for the same primary endpoint. The proposed method is illustrated using a case study consisting of 3595 candidate biomarkers for a bioactivity primary endpoint.
|