Explanatory epidemiological models from genotype to phenotype

There is evidence that genes implicated in similar diseases reflect underlying physiological networks, such that mutations in genes that normally interact with each other result in similar diseases. Common diseases, such as depression or diabetes result from having multiple genes affected in the same individual; the challenge is to identify which genes and variants are important and how the combination affects risk. We will develop a computational modelling approach examining diseases which occur together in the same individual (co-morbidity), and will identify common or overlapping disease mechanisms across the genome through correlating allele frequencies with observed relative risk. If successful, our research will identify molecular mechanisms for a large number of genetically-based complex diseases. Understanding them will aid in improving diagnosis, treatment, and prevention, and will have significant public health benefits. Our work consists of two steps, data preparation and pre-processing, and model construction. In the first step, we will extract relevant features from the genomic and phenotypic information in the cohort, and perform standard statistical analyses for genetic associations with disease, relative risk between diseases, and frequency of genetic variants within the cohort. In the second step, we will build a computational model (an algorithm) that explains the observed co-morbidities through the observed genetic variant frequencies. We will utilize all available genotype and phenotype data, i.e., the full cohort. This is to ensure the generality of our approach and to obtain sufficient statistical power for our statistical model.