This Small Business Innovation Research project addresses the problem of biomarker detection in clinical and high-throughput data. The objective is to investigate new approaches for deter- mining, from data consisting of many possibly irrelevant or redundant measurements, a highly predictive and interpretable model that involves only a small number of measurements. These new methods will be studied for modeling subjects'time-to-event (such as stroke, heart attack, or metastasis in cancer). The proposed approaches will be compared with existing methods that attempt to use relatively few mea- surements in modeling survival (time-to-event) data. The data to be analyzed will include ion-mobility and clinical data from a large cardiovascular disease cohort, as well as high-throughput genomic data from cancer research with many more measurements than samples. Relevance. Although today's advanced technologies offer the possibility of revolutionizing clinical practice, the analytical tools available for extracting information from this amount of daa are not yet sufficiently developed for targeted exploration of the underlying biology. This project directly addresses the need to make what the FDA terms IVDMIA (In-Vitro Diagnostic Multivariate Index Assays) transparent and interpretable, and is thus an opportunity to improve analysis services or products provided to companies that identify, characterize, and validate biomarkers for clinical diagnostics and drug development decision points. The proposed project will produce robust methods for parsimonious biomarker detection that will speed the development of cheaper and more effective diagnostic tests for disease diagnosis, treatment monitoring, and therapeutic drug development.
There is a great need in medical research for prognostic models that can accurately predict time to an event, such as a heart attack, from a few observed features. These models can be used in establishing new diagnostic and screening tests, and in advancing new therapies. New methods for time-to-event modeling are proposed that will speed the development of cheaper and more effective clinical support systems, and have a far-reaching impact on public health.