Christoph Lange - US grants

The Biostatistics Core will be responsible for statistical aspects of experimental design and data analysis, and the development of new statistical methods required by Projects 1 and 2. In collaboration with the Bioinformatics Core, the Biostatistics Core will design and implement analysis methods and associated software for use by investigators in the analysis of resequencing and genotype data from the Projects. The specific aims are: 1) design and implement an integrated approach for data analysis to support the day to day data analysis demands of the Projects by developing standardized protocols for the statistical analysis of both family-based association studies and case-control studies to maintain adequate statistical power while providing appropriate control of Type I error in the face of a potentially large number of tests;2) develop novel statistical genetics methods and software to meet the specific needs of each of the Projects and the other Cores and in collaboration with the Bioinformatics Core, make these new implementations available to the research community through the public web interfaces provided by the Bioinformatics Core. The Biostatistics Core will be directed by Dr Christoph Lange and co-directed by Dr Nan Laird. Drs Lange and Laird are world-renowned statisticians with extensive experience in the development of important new methodology for statistical genetic analyses, developing and distributing suites of data analysis software, and in the analysis of genetic data sets and disease association studies. They have a well-established track record of highly productive collaboration with each other and with the other investigators assembled for this project. The work of the Biostatistics core will contribute to our understanding of genetic and environmental influences in asthma and COPD, two common and important human diseases which display complex patterns of genetic influences and important effects from environmental exposures. In addition, the work will lead to improvements in statistical genetics methods for complex traits.

DESCRIPTION (provided by applicant): The immanent influx of high-throughout sequencing datasets poses both a unique opportunity to identify the disease susceptibility loci for complex disease and their pathways and a challenge in terms of the statistical analysis. Many of the loci that are recorded by high-throughput sequencing studies will be rare, providing insufficient power for the statistical analysis. For studies with unrelated cases and controls, a number of collapsing approaches has been suggested. However, such methodology does not exist for family-based studies which are by design well suited for rare-variant analysis. They have higher statistical power for rare variants and are robust against population admixture. For population-based designs, statistical approaches that adjust the analysis for such confounding do not exist if the variants are rare. However, for the construction of collapsing method for family-based designs, the linkage disequilibrium (LD) between the loci has to be estimated which is a non-trivial task for rare variants. In population-base designs, this issue can be avoid by utilizing permutation tests that randomly assign the phenotype, but keep the genetic data in a subject fixed. This is not possible in family-based designs. In this grant application, we will develop an analytical approach to the LD-estimation problem in family-based designs. This will enable the construction of rare variant tests for family-based designs. The major goal of sequence-analysis is the identification of the DSLs. The significance of single-locus association tests is defined by the genetic effect size and the allele frequency. Since non-DSLs that are in LD with the true DSL can have higher allele frequencies than the DSL, but have smaller, observed genetic effect sizes, the significance of the test cannot be used to identify DSLs. In order to distinguish the true DSLs from SNPs that are in LD with the DSLs, we will develop statistical approaches that assess differences in LD-pattern across multiple loci between subjects are required. Such methodology will be proposed for designs of unrelated individuals and family-based studies. The new analysis approaches will be integrated in our software packages. The new approaches will support the search for disease loci in the human genome which will lead to a better understanding of the pathways for complex diseases and ultimately to their treatment. PUBLIC HEALTH RELEVANCE: Sequencing data contains the information that is needed to identify the causal genetic loci for complex diseases and phenotypes. However, to translate this wealth of information into the discovery of disease loci, novel statistical analysis approaches are required. While the current analysis methodology remains valid, they are not optimally designed to look at rare variants and sequence data. We will develop statistical tools that are robust against confounding in rare variant data and that can identify the locations of the disease loci in sequencing data. This important information will support the search for disease pathways and their cure.

DESCRIPTION (provided by investigator): Genome-wide association studies (GWAs) have led to the discovery of novel, robust associations for numerous complex diseases and phenotypes. While the new findings can be replicated reliably in other studies, the amount of phenotypic variation that is explained by the new association findings is small compared to the estimated total heritability of most diseases/traits. This suggests that the current GWAs are not able to identify most of the disease loci. Potential reasons are the study heterogeneity/confounding and the lack of sufficient statistical power to address the inherent multiple testing problem. For family-based designs, we will develop novel statistical methodology that achieves higher power levels than the currently used methodology and, at the same time, are completely robust again confounding. The application of the new methods to genome-wide association studies for Alzheimer's'Disease and Attention Deficit Hyperactivity Disorder will provide new insights that will help the scientific community to identify new genes for these diseases which are major public health problems in the United States. PUBLIC HEALTH RELEVANCE: Alzheimer's disease and Attention Deficit Hyperactivity Disorder are major public health problems in the United States. The proposed statistical methodology will provide new analysis approaches that will enable researchers and clinicians to identify genetic risk loci for these diseases and other complex disease and phenotypes. In turn, an improved understanding of the genetic architecture of these conditions will result in a better and more efficient care for those who suffer from these diseases.

ABSTRACT The overarching goal of the program project, ?Systems Biology of Airway Disease?, is to identify common molecular determinants and pathways for asthma and COPD. These determinants will be identified through the use of a diverse array of molecular data ? DNA sequencing and genomewide SNP data (Project 1), RNA sequencing and expression data (Project 2), methylation sequencing and miRNA sequencing data (Project 3). In order to meet the goals of this project, extensive biostatistical and bioinformatics support is necessary and will be implemented throughout each of the three projects. As such, we have established a Biostatistics and Bioinformatics Core (Core B) that will meet these needs for the program project. Specifically, we will develop/refine bioinformatics pipelines for DNA, RNA and methylation sequencing data to ensure that the sequencing data to be generated will be of high quality and are managed properly. These pipelines will be incorporated into the existing bioinformatics structure that supports the other molecular data being used throughout this PPG. Several of the goals throughout the three projects are innovative in nature and are best addressed through the development of new statistical methods that are specific to the project aims. As such, this core will also develop and distribute novel statistical genetics methods and software necessary to meet the specific needs of each of the projects. Finally, the goals for each project require extensive statistical analyses. This core will oversee the statistical analyses throughout the PPG, ensuring that the analyses are appropriate and completed in a timely manner. The Biostatistics and Bioinformatics Core will be directed by Dr. Christoph Lange, Professor of Biostatistics at the Harvard School of Public Health (HSPH), with a joint appointment as Assistant Professor of Medicine at the Harvard Medical School (HMS). In addition, there is a team of experienced statisticians, bioinformaticists and bioinformaticians who are longtime collaborators with each other and with the Project Leaders, Dr. Scott T. Weiss (Project 1), Dr. Benjamin A. Raby (Project 2) and Dr. Dawn L. DeMeo (Project 3). Their successful track record and productivity suggests that the Biostatistics and Bioinformatics team will work successfully to help meet the overall needs of this PPG.

? DESCRIPTION (provided by applicant): The availability of next-generation sequencing data in large-scale association studies provides a unique research opportunity. The data contains the information that is required to identify causal disease susceptibility loci (DSL) for many mental health phenotypes and psychiatric diseases. In order to translate the wealth of information into DSL discovery, powerful statistical methodology is required. So far, a large number of rare variant association tests have been proposed. However, they do not incorporate all the important information about the variants. So far, none of the existing approaches takes the physical location of the variant into account. Under the assumption that deleterious DSLs and protective DSLs cluster in different genomic regions, we will develop a general association analysis framework that is built on spatial clustering approaches. The framework will be able to handle complex phenotypes, e.g. binary, quantitative, etc., and be applicable to different study designs, i.e. family-based studies and designs of unrelated subjects. If the DSLs cluster indeed, the increase of statistical power of the approach will be of practical relevance, enabling the discovery of DSLs. In the absence of DSL clustering, our approach will achieve similar power levels as existing methodology. Furthermore, in order to test larger genomic regions for association, we will develop network-based association methodology. The network-based approach will have sufficient power for larger genomic region than existing approaches, and, at the same time, provide an intuitive understanding of the complex relationships between the variants that drive the association, fostering new biological insights. The approach can incorporate complex phenotypes and different design types. We will also use the information about the physical locations of the rare variants to detect population substructure/admixture. Since rare variants are genetically much younger than common variants, approaches that take the physical locations of the variants and their clustering into account will provide a much finer resolution picture of population substructure in sequence data than existing approaches, e.g., EIGENSTRAT. We will use community-detection algorithm for the classification of study subjects in genetic homogenously subgroups. All the proposed methodology will be implemented in user- friendly software packages with existing user-communities, i.e. PBAT, NPBAT and R.