Integrative data analysis for biomarker discovery
This chapter focuses on the combination of different types of information and prediction models for biomarker discovery. The importance and application of integrating different types of data and computational approaches will be discussed. Methodologies and tools for integrating and analyzing different data sources will be introduced. Examples of approaches to supporting the identification of biomarkers for disease classification and the prediction of clinical outcomes will be provided.
The combination of multiple biomarkers derived from different clinical and molecular data sources have been proposed to improve diagnostic and prognostic performance in different research areas, especially in cancer and cardiovascular research. For example, traditional risk factors (such as age, gender and blood glucose concentration) have been combined with protein expression biomarkers (such as different molecules implicated in inflammation processes) to improve the prediction of recurrent cardiovascular events in comparison to traditional risk factors (Blankenberg et al., 2006).
In a traditional integrative approach to disease prediction model design, the different biomarkers are independently discovered prior to the integrative modelling task. Typically they are selected as inputs to these analyses based on prior knowledge, that is new models are investigated based on the combination of known traditional risk factors or novel ...