Sponsoring NCI Division:	DCCPS
		Grant Number:	R37 CA076404-11
		Award Approved:	June 2007
		Institution:	Harvard School of Public Health, MA
		Department:	Biostatistics
		Scientific Abstract (CRISP) Xihong Lin Xihong Lin, Department of Biostatistics Literature Search on PubMed

Statistical Methods for Correlated and High-Dimensional Biomedical Data

In the research supported by this MERIT award, Dr. Lin will develop advanced statistical methods for analyzing correlated and high-dimensional data in cancer research, especially for analyzing longitudinal and familial data, and high-dimensional genomic and proteomic data in epidemiological studies and population sciences.

An important area in genomic epidemiological studies is the need to select Single Nucleotide Polymorphisms (SNPs), genes, or gene (SNP)/protein sets, e.g., genetic pathways, from a large pool of genes (SNPs) that are related to disease outcomes. Dr. Lin will develop advanced statistical informatics methods for high-dimensional genomic data in population studies, such as gene (SNP) selection, joint effects of genes in a genetic (metabolic) pathway, gene-gene (SNP-SNP) interactions, and gene-environment interactions, while accounting for the observational nature of epidemiological studies. Such methodological developments will proceed in close collaboration with basic scientists, genomic epidemiologists, and clinicians at Harvard School of Public Health, Dana-Farber Cancer Institute, and Massachusetts General Hospital on large scale genomic cohort and case-control studies, especially in lung cancer, the leading cause of cancer deaths.

Dr. Lin will also develop advanced statistical informatics methods for analyzing proteomic mass spectrometry (MS) data. The emerging field of comparative proteomics presents an exciting opportunity for new cancer biomarker discovery. However, analysis of high-dimensional proteomic MS data is challenging. Dr. Lin will develop advanced functional learning methods for properly pre-processing high-dimensional proteomic MS data and for selecting proteins (peptides) as new biomarkers, studying protein-protein (peptide-peptide) interactions and protein-environment interactions, while accounting for the observational nature of population studies. The methodological development will be in close collaboration with Harvard/Dana-Farber proteomic researchers in basic science and population studies.

Software development would be an important component of the MERIT award so that these advanced statistical methods can be readily available to genomic and proteomic researchers in population sciences.