Pani: a novel algorithm for fast discovery of putative target nodes in signaling networks

Abstract

In biological network analysis, the goal of the target identification problem is to predict molecule to inhibit (or activate) to achieve optimum efficacy and safety for a disease treatment. A related problem is the target prioritization problem which predicts a subset of molecules in a given diseaserelated network which contains successful drug targets with highest probability. Sensitivity analysis prioritizes targets in a dynamic network model using principled criteria, but fails to penalize off-target effects, and does not scale for large networks. We describe Pani (Putative TArget Nodes PrIoritization), a novel method that prunes and ranks the possible target nodes by exploiting concentration-time profiles and network structure (topological) information. Pani and two sensitivity analysis methods were applied to three signaling networks, mapk-pi3k; myosin light chain (mlc) phosphorylation and sea urchin endomesoderm gene regulatory network which are implicated for example in ovarian cancer; atrial fibrillation and deformed embryos. Predicted targets were compared against the molecules known to be targeted by drugs in clinical use for the respective diseases. Pani is orders of magnitude faster and prioritizes the majority of known targets higher than both sensitivity methods. This highlights a potential disagreement between absolute mathematical sensitivity and our intuition of influence. We conclude that empirical, structural methods like Pani, which demand almost no run time, offer benefits not available from quantitative simulation and sensitivity analysis. Copyright © 2011 ACM.