New Techniques to Combine Measures of Statistical Significance from Heterogeneous Data Sources with Application to Analysis of Genomic Data

This project is motivated by integrative analysis of large-scale genomic data, where an important question is how to effectively combine statistical significances, or p-values, from heterogeneous data sources. Despite recent advances in theoretical and applied studies, statistical and computational challenges remain in addressing critical data features, such as complex correlations, discreteness of data, and availability of prior knowledge that could have been utilized to boost signal detection.

This project will develop novel statistical methods to address the challenges and increase the statistical power for detecting valid signals. The research will facilitate innovations in statistical theory and methodology as well as in broad applications. The research activities will leverage project-oriented education, promote multi-disciplinary interactions, and benefit STEM education for the next generation of engineers and scientists, especially members of minorities underrepresented in the statistics field.

Specifically, the project will develop efficient and powerful p-value combination tests by following a new strategy different from common literature. Instead of designing and studying tests individually, the project will strategically resolve problems based on general families of tests. The project has three specific research aims.

The first aim is to tackle the computational challenges in applying the p-value combination approach into analyzing complex heterogeneous data. The PI will develop fast and accurate algorithms to control the error rates of general families of tests under general correlations and the discreteness of the p-values. The second aim is to increase the power of p-value combination for integrative analysis of complex data through utilizing the correlation information, incorporating prior knowledge, automatically adapting to complementary procedures, and revealing asymptotic optimality properties.

Finally, the PI will apply the developed methods into the integrative analysis of large-scale genomic data of neurodegenerative diseases. Results of the project are expected to advance global hypothesis testing methods for high-dimensional complex data analysis.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.