Modern biotechnologies have revolutionized human genomics, allowing researchers to query across the ~3.2
billion nucleotide base positions that form the human genome. The ability to detect genetic variants in a high-
throughput manner has revealed specific alleles that contribute to human phenotypic variation, including those
associated with disease risk. However, the vast majority of studies have focus exclusively on the portion of the
genome located in the nucleus. Largely ignored is the DNA contained in the cell's mitochondria (mtDNA), which
harbors the genes encoding proteins that are responsible for generating most of the cell's energy. Importantly,
the large and variable numbers of mitochondrial chromosome copies per cell can result in an mtDNA variant co-
existing with a wild-type allele, a condition known as heteroplasmy. The variant's heteroplasmy level can shift
dramatically across generations, as well as spatially/temporally within the same individual. With new data
sources, it is possible for the first time to develop highly accurate models of the processes underlying
mitochondrial genome dynamics. Since these processes have hierarchical aspects, Bayesian hierarchical
modeling is an ideal framework within which to develop such models.
Given this, we propose to pursue the following three Specific Aims: 1) Develop a flexible Bayesian modeling
framework to capture mtDNA dynamics; 2) Apply the framework to large data sets from a variety of clinically-
relevant settings; and 3) Comprehensive model testing, experimental validation, and implementation.
Mitochondrial DNA mutations have been implicated in disease phenotypes including diabetes, autism,
encephalomyopathies, stroke, vision loss, cancer, and many others. As additional relevant data become
available, further elucidation of the impact of mtDNA variation on complex phenotypes is possible. Such insights
are likely to lead to important discoveries in genetics as well as medical applications. This project will facilitate
advances by providing reliable computational infrastructure for efficient and accurate modeling of mtDNA
mutational dynamics. The resulting software will be implemented and disseminated in the free and portable
statistical computing platform R.