Recognition of gene regulatory sequences by bagging of neural networks

The authors use an ensemble of multilayer perceptrons to build a model for a type of gene regulatory sequence called a G-box. A variant of the bagging method (bootstrap-and-aggregate) improves the performance of the ensemble over that of a single network. Through a decomposition of the generalization error of the ensemble into bias and variance components, the authors estimate this error from the hold-out samples of the individual networks. They test the model on putative G-boxes, on sequences upstream of light-regulated genes, and on a control group and demonstrate that the model separates these groups efficiently.