Genomic Network Tomography

This paper considers the problem of learning cellular signaling networks from incomplete measurements of pathway activity. Cells respond to environmental changes (e.g., starvation, heat shock) via a sequence of intracellular protein-protein interactions, leading to the production of proteins which modify their fundamental operations. Biologists have discovered some of these signaling pathways, but the knowledge of cellular signaling is still very incomplete. Mathematically, the problem of genomic network tomography (GNT) - identifying cellular signaling networks from biological data - is similar to network inference problems arising in communication systems. This paper formulates GNT and presents a solution which builds on state-of-the-art communication network inference techniques while taking into account uncertainties which are inherent in biological data.