A schematic showing the problem of finding PAGs, existing methods to solve the problem, and the principle of our solution. (a) A compound alters the expression of PAGs in a cell. The change in gene expression then propagates to other genes through the gene network, but genome wide expression analysis only shows a number of expression changes. Thus PAGs cannot be distinguished from their secondary effects from the micro-array data. The CutTree algorithm ranks the genes according to micro-array experiments and identifies the PAGs. (b) To find drug targets, different strategies can be used. With a network identification method (left), the structure and dynamics of the underlying cellular network is identified, and the targets of a compound can be calculated. A random search (middle) can be performed by extending the expression library until the knockout genes coincides with the targets. CutTree (right) infers multiple targets using a small number of experiments.