Identifying Responsive Modules by Mathematical Programming: An Application to Budding Yeast Cell Cycle

Abstract

High-throughput biological data offer an unprecedented opportunity to fully characterize biological processes. However, how to extract meaningful biological information from these datasets is a significant challenge. Recently, pathway-based analysis has gained much progress in identifying biomarkers for some phenotypes. Nevertheless, these so-called pathway-based methods are mainly individual-gene-based or molecule-complex-based analyses. In this paper, we developed a novel module-based method to reveal causal or dependent relations between network modules and biological phenotypes by integrating both gene expression data and protein-protein interaction network. Specifically, we first formulated the identification problem of the responsive modules underlying biological phenotypes as a mathematical programming model by exploiting phenotype difference, which can also be viewed as a multi-classification problem. Then, we applied it to study cell-cycle process of budding yeast from microarray data based on our biological experiments, and identified important phenotype- and transition-based responsive modules for different stages of cell-cycle process. The resulting responsive modules provide new insight into the regulation mechanisms of cell-cycle process from a network viewpoint. Moreover, the identification of transition modules provides a new way to study dynamical processes at a functional module level. In particular, we found that the dysfunction of a well-known module and two new modules may directly result in cell cycle arresting at S phase. In addition to our biological experiments, the identified responsive modules were also validated by two independent datasets on budding yeast cell cycle.

Quantitative assessment of gene expression network module-validation methods

“…For this reason, SBA is an important process to assess a module’s stability, phenotypic correlation or significance of consistency ( Table 1 ). For responsive modules or module biomarker identification 50 51 52 , binary or mixed integer linear programming models can be used to validate the causal or dependent relations between network modules and biological phenotypes 34 53 54 . In phylogeny, resampling approaches are defined as a confidence measure for splits in a phylogenetic tree and are used to calculate consensus trees 55 , which can also be used to assess the robustness of modules in network analysis 25 56 .…”

Network-based drug repositioning

“…Guided by this framework, we also proposed several computational methods to detect responsive networks from PPI networks by leveraging microarray data. 20,[65][66][67][68][69] Based on global characteristics of interactome coupled with gene expression data, we developed a novel method to detect disease-related gene modules or dysfunctional pathways. In this method, interactions among genes are exploited to define a gene's activity score and responsive networks are inferred by the support vector regression.…”

Genetic interactions derived from high-throughput phenotyping of 6589 yeast cell cycle mutants

“…Because of the intense labor involved in these experiments, individual laboratories have tended to focus on small numbers of genes and proteins involved in subsections of the extensive network of gene/ protein interactions that control cell cycle events. This reductionist approach was necessary in the early stages of identifying and characterizing the molecular regulatory system, but it carries with it the danger of missing higher levels of network organization and their phenotypic consequences [2][3][4] .…”

“…To mitigate these problems, many researchers, including ourselves, have developed detailed mathematical models that integrate top-down and bottom-up approaches in order to describe the molecular mechanisms that underlie cell cycle regulation in budding yeast 4,[17][18][19][20][21][22] . The governing equations of the model are simulated on a computer, and the model (the 'wiring diagram' of molecular interactions) is adjusted until it generates dynamic behaviors that reflect the documented molecular changes and general network behaviors observed in cells (e.g., cell viability, timing of cell cycle events, cell size at birth, and response to DNA damage or chromosome misalignment at mitosis) [23][24][25][26] .…”