Gene hunting with forests for multigroup time course data

Gene hunting with forests is a new method for identifying differential gene expression profiles across experimental groups using time course data. Our approach utilizes a multi-dimensional filter that captures the functional nature of the data while adjusting for additional variables that may be part of the experimental design. The filter comprises one component measuring gene profile differences, and another component measuring estimation error. Interesting genes are those having substantial gene profile differences and low estimation error. We refer to this as our Gene Hunting Principle. We illustrate this methodology using a balanced design, involving the effects of muscle group-specific gene expressions on postnatal development. We also consider a more complex experimental design focusing on the effects of aging in the human kidney.