Figure 1
From: Network-based de-noising improves prediction from microarray data
De-noising based on a network. In the top figure, the target expression vector to be de-noised is depicted as a red point. Black points are the neighbors in the network (below) derived from side information; gray points are vectors which are not directly connected (i.e. related) to the target vector in the network. The edge of the network is depicted by a solid line. A dashed curve indicates the correspondence between data in a network and the expression vector. De-noising is done by robust projection onto the principal subspace made from only the neighbors. In this case, the subspace (gray line) is obtained by PCA of the target and the four neighbors.