I am a biologist trained in statistics, evolutionary genetics, morphometrics, and statistical genomics. In my work I seek to understand the genetic and developmental causes of phenotypic variation from an evolutionary standpoint. What follows summarizes what this means in a bit more detail. On the right, I've placed links to descriptions of specific projects I work and have worked on.

My research asks about the mechanisms that are responsible for phenotypic variation, that inescapable force of nature that results from not two iterations of the same developmental program yielding quite the same results. I am particularly interested in how variation of highdimensional systems is mapped between causal and target spaces (e.g., between the space of DNA sequence variation and the space of variation in gene expression). Here "mapping," a term borrowed from topology, describes a quantitative representationa modelof a hypothesis of causation; more generally, mapping is the cognitive process through which we link chains of evidence to postulate a causeeffect relationship. Mapping underlies regression and correlation statistics, and it is central to all forms of experimentation.
The complexity of biological mappings varies widely; from simple experiments with one predictor and one response variable, to problems where both causal and target spaces are highdimensional. The genotypephenotype map, which is actually a collection of many submappings, belongs in the second category. The complexity that results from high dimensionality has tremendous practical implicationsit can make it very difficult to predict the precise response of a system to even simple stimuli, let alone responses to highdimensional, poorly understood stimuli.
The overarching aim of my work has been, since my undergraduate years, to try and make sense of the network of causes underlying complex patterns of phenotypic variation. Consistent with most biomedical and evolutionary research, the guiding aim of my work is to uncover sets of rules and conditions that allow us to predict genetic and phenotypic responses to change, whether such change is brought about by internal (e.g., mutation) or external (e.g., environmental) factors. My research has and continues to be benefited by the incorporation of finer genomic and phenomic data and increasingly sophisticated mathematical approaches to deal with the nonEuclidean, highdimensional spaces where phenotypic data usually reside.
In my current research on Drosophila melanogaster, my colleagues and I are using DNA sequence data and gene expression perturbation experiments to find genetic correlates of phenotypic variation in the shape of the wing. Both of these studies treat shape as a multidimensional trait, which has allowed us to move beyond the mere listing of candidate genes with a phenotypic effect to focus on the structure of the effects themselves. This is because multivariate analyses such as MANOVA estimate the directions of variation that maximize discriminating power based on the combined information of all of the dimensions measured in a sample. We are exploiting these directions in several ways, one of which being the construction of phenotypic networks that point to the postulation of candidate gene interactions. In the process, we have become convinced that not only are multivariate approaches to genomewide studies more informative, but also they might be the most costefficient way of increasing power to determine causal factors of variation in complex phenotypes.
In the future, I expect to steer my research toward the study of the determinants of susceptibility and pathogenesis in complex human diseases (e.g., MetS). I have found that exposure to the complexities of statistical analysis of shape has been excellent preparation to confront the level of complexity required by genotypephenotype mapping from a systems biology perspective. Given the deluge of phenomic data coming our way, understanding the causes of phenotypic variation looks to be one of the greatest challenges we will be facing in the upcoming decades, as we will need to come to terms with the high level of integration, nonlinearity, and dimensionality that pervades most aspects of the phenotype.

Present and Past Projects
