There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Nicholas Schork |
| Institution: | The Translational Genomics Research Institute |
| Department: | Quantitative Medicine |
| Country: | |
| Proposed Analysis: | Heterogeneity, or the phenomenon whereby subsets of individuals express a phenotype for different reasons, is a biological reality and often confounds genetic and biomarker studies. There are many forms for heterogeneity. For example, ‘allelic heterogeneity’ occurs when different variants in the same gene disrupt that gene in similar ways and lead to disease; ‘locus heterogeneity’ occurs when different genes contribute to a disease independently of one another (i.e., different genes could be sufficient but not necessary for a disease); ‘etiologic heterogeneity’ occurs when different genetic networks or genetically-mediated processes contribute to a disease; and ‘clinical heterogeneity’ occurs when similar genetic and etiologic processes are perturbed but lead to different clinical manifestations. Identifying and controlling for different forms of heterogeneity in genetic analyses is not trivial. One way to reveal heterogeneity is to determine if there are, e.g., clinically relevant phenotypic differences between individuals based on known or measured factors collected on those individuals, such as sex or age or ancestry (e.g., is blood pressure different between men and women?). If evidence for another factor, such as a genetic variant, is found whose relationship with the relevant clinical phenotype is of interest, then one could determine if the relationship between that variant and the clinical phenotype is modified by (or interacts with) another measured factor, like sex. If it is, then one could say that there is heterogeneity in the way the genetic factor influences the clinical phenotype and sex can explain that heterogeneity. When one does not know of specific factors that could modify the relationship between a genetic factor and a phenotype, then one could either search for such factors (which could involve quite a large search with many statistical problems likely to arise, most notably the potential for false positive and negative findings) or simply posit a parameter in relevant models relating the genetic factor to the phenotype that accounts for heterogeneity, whatever its origins. Models that account for heterogeneity in the way a genetic variant influences a phenotype that do not explicitly assume that a particular factor is responsible for the heterogeneity can be developed using mixture models, of which I have had experience developing. Mixture models include a parameter which quantifies the proportion of observations that, e.g., do not follow a particular trend or pattern that other observations do. This parameter, known as the mixing parameter, can be estimated using various techniques, such as maximum likelihood techniques, but invites some thorny statistical problems that can be overcome using simulation-based inference techniques. I will use mixture models to: 1. explore genetic associations (i.e., explore allelic and locus heterogeneity); 2. test for potential causal relationships between genetic factors, intermediate phenotypes and overt clinical phenotypes (i.e., test for etiologic heterogeneity), and 3. Seek to identify genetic factors that influence disease processes impacting more than one clinical phenotype (i.e., seek to identify clinical heterogeneity). |
| Additional Investigators |
