Development of statistical methodologies and application to clinical cancer studies (2015–2017)
Abstract:
The pace of technological innovation in high throughput biology combined with decreasing costs affords researchers with improved scale, depth and resolution of data. However, current methodological approaches are limited, and do not always fulfill the potential of these studies. This project will develop innovative statistical solutions for the identification of biomarkers and molecular targets in complex designs and settings and apply these to key cancer research project as the skin and blood cancer programs led in my institute. I have identified several typical
research questions in clinical and cancer studies, such as: what is the ulti-marker panel a) which can distinguish patients outcomes across multiple `omics¿ experiments, b) which can predict the progression of a disease or decipher functional dynamics, c) which is robust and reproducible across related experiments performed on different sets of patients and in different labs? These questions are related to the integration of disparate types of biological data measured on different platforms, across time points (longitudinal experiments) or across different studies (cross-platform comparison). The methods I propose to develop will be based on multivariate statistical approaches such as Partial Least Squares regression variants in which I have considerable experience. They will be applied to some key studies from my institute and will be publicly released in the form of a software for a broad range of biological systems where data integration is required. Through this project, my goal is to bridge the gap between data generation in the lab and data validation in the clinic.
