Physical Network Models

Abstract

We develop a new framework for inferring models of transcriptional regulation. The models, which we call physical network models, are annotated molecular interaction graphs. The attributes in the model correspond to verifiable properties of the underlying biological system such as the existence of protein-protein and protein-DNA interactions, the directionality of signal transduction in protein-protein interactions, as well as signs of the immediate effects of these interactions. Possible configurations of these variables are constrained by the available data sources. Some of the data sources, such as factor-binding data, involve measurements that are directly tied to the variables in the model. Other sources, such as gene knock-outs, are functional in nature and provide only indirect evidence about the variables. We associate each observed knock-out effect in the deletion mutant data with a set of causal paths (molecular cascades) that could in principle explain the effect, resulting in aggregate constraints about the physical variables in the model. The most likely settings of all the variables, specifying the most likely graph annotations, are found by a recursive application of the max-product algorithm. By testing our approach on datasets related to the pheromone response pathway in S. cerevisiae, we demonstrate that the resulting model is consistent with previous studies about the pathway. Moreover, we successfully predict gene knock-out effects with a high degree of accuracy in a cross-validation setting. When applying this approach genome-wide, we extract submodels consistent with previous studies. The approach can be readily extended to other data sources or to facilitate automated experimental design.

Large-Scale Elucidation of Drug Response Pathways in Humans

“…Accordingly, we inferred for each drug the most likely subnetwork connecting its targets to its induced set of differentially expressed genes, computed from drug response gene expression profiles extracted from the Connectivity Map database (Lamb, 2007). The inference aimed at maximizing the likelihood of the resulting subnetwork while minimizing the lengths of its constituent pathways (see Methods); following the work of Yeang et al (2004), each such pathway, connecting a drug target to a differentially expressed gene, was required to end with a protein-DNA interaction (which mediates the effect of the drug on the expression of the target gene). Overall, we obtained 428 drug-specific subnetworks with an average size of 19 proteins per subnetwork.…”

The Approximability of Shortest Path-Based Graph Orientations of Protein–Protein Interaction Networks

“…Identifying this directionality is fundamental to our understanding of how these protein networks function. To this end, previous work has relied on information from perturbation experiments (Yeang et al, 2004), in which a gene is perturbed (cause) and, as a result, other genes change their expression levels (effects). The fundamental assumption is that, for an effect to take place, there must be a directed path in the network from the causal gene to the affected gene.…”

“…Other approaches to the problem concentrated on short connecting paths, which are more plausible biologically (Yeang et al, 2004). Gitter et al (2011) focused on paths whose length is bounded by a parameter k, showing that although the resulting problem is NP-hard, it can still be approximated to within a factor of O(2 k /k).…”

Application of a systems approach to study developmental gene regulation

“…Many algorithms have been adapted to make use of these physical and causal interaction information jointly for inferring gene regulatory network. One proposal was to construct a physical network model to connect functional perturbation experimental results with a physical network of protein-protein and protein-DNA interactions (Yeang et al 2004). In the initial study, the physical network was used to define all possible edges between nodes, and the functional perturbation data was used to select the active edges and assign edge direction using a probabilistic graphical model.…”