In theory, selected gene knockouts should be sufficient to characterize a biochemical pathway however in practice this characterization can be difficult. In this talk I present a method for automated pathway inference from gene knockout data to identify both novel pathways and targets for this pathway.

The examples will focus on detecting variants of the developmental pathway governed by the protein Sonic Hedgehog (Shh). In mammals, Shh is in part responsible for limb and brain development. Errors in Shh signaling result in severe developmental defects in the embryo or a wide variety of cancers in adults.

When applied to gene knockout data based on a series of gene knockouts, this pathway identification approach strongly suggests at least one additional mechanism for Shh signaling that has not been previously identified. As a byproduct, the algorithm also generates a list of highly probable targets for each of these pathways. These detected target genes include most well-known Shh targets and also novel targets not previously associated with Shh signaling.

This work is supported by the NIH