Genome‐scale predictions and profiles of mitochondrial function

Abstract

Mitochondria are tiny organelles found in nearly all cells, serving as the center stage for ATP production, ion homeostasis, and apoptosis. Their molecular composition, structure, and efficiency are dynamic, varying across cell types and adapting to environmental cues. Mitochondrial dysfunction can contribute to numerous, rare metabolic disorders as well as to a variety of common human diseases, such as diabetes and cancer. With the goal of systematically exploring mitochondrial function in health and in disease, we are using experimental and computational genomics to (1) assemble a high quality protein parts list for this organelle and to (2) monitor their collective behavior. We currently estimate that nearly 1500 nuclear genes encode protein components of human mitochondria. At present, only one-half are known. To predict the remaining genes, we are generating and integrating genome-scale datasets using machine learning techniques. Our methods achieve high sensitivity and specificity and identify hundreds of novel mitochondrial genes, including novel candidate disease genes. We are exploring their collective behavior by monitoring their expression with oligonucleotide microarrays. One of our key findings is that the expression of nuclear mitochondrial genes is coordinately reduced in the muscle of individuals with type 2 diabetes mellitus. We’ve developed new computational strategies that combine expression analysis with comparative sequence analysis to dissect cis-regulatory networks. Using these techniques, we’ve identified two key transcription factors that form a regulatory circuit that controls mitochondrial biogenesis in skeletal muscle. Importantly, the results identify a valuable new drug for type 2 diabetes.