Systems and methods for modeling and analyzing networks

Abstract

The systems and methods described herein utilize a probabilistic modeling framework for reverse engineering an ensemble of causal models, from data and then forward simulating the ensemble of models to analyze and predict the behavior of the network. In certain embodiments, the systems and methods described herein include data-driven techniques for developing causal models for biological networks. Causal network models include computational representations of the causal relationships between independent variables such as a compound of interest and dependent variables such as measured DNA alterations, changes in mRNA, protein, and metabolites to phenotypic readouts of efficacy and toxicity.

What is claimed is: 
     
         1 . A method of building a computer model to extract information from a dataset comprised of two or more variables, comprising (i) inferring said computer model containing equations describing the relationships between said variables, (ii) simulating said computer model to predict the impact of the change made to the value of one or more first variables on the values of one or more second variables. 
     
     
         2 . A method as in  claim 1 , wherein the inferring step is carried out as follows: (a) building local models by (i) selecting a set of interaction forms to define the quantitative relationships between variables in said local models (ii) building local models by proposing connections between two or more of said variables and using a scoring method to determine how likely such local models are given the data (iii) creating a library of local models ranked according to a score generated by said scoring method; (b) building global models by choosing local models from said library of local models and connecting said local models. 
     
     
         3 . A method as in  claim 2  wherein the building of local models is achieved by a global optimization method. 
     
     
         4 . A method as in  claim 2  wherein the global optimization method is metropolis Monte Carlo. 
     
     
         5 . A method as in  claim 1  wherein the search space to be searched in the inferring step is constrained using prior information about the variables in the dataset. 
     
     
         6 . A method as in  claim 1  wherein the values of some or all of the variables in said model are displayed on top of or next to their corresponding representation in a graphical depiction of said model. 
     
     
         7 . A method as in  claim 6  wherein the graphical depiction is a directed acyclic graph. 
     
     
         8 . A method as in  claim 1  wherein the reverse-engineered model is represented using Diagrammatic Cell Language. 
     
     
         9 . A method as in  claim 1  wherein the model created is a consensus model comprised of two or more underlying models that together reflect the process that gave rise to the dataset. 
     
     
         10 . A method as in  claim 1  wherein the model created contains variables reflecting two or more types of measurements. 
     
     
         11 . A method as in  claim 1  wherein the simulation comprises implementation of a computer script to automatically change the value of one or more said first variables and record or display the resulting values of one or more second variables in the simulation. 
     
     
         12 . A method as in  claim 1 , wherein the information to be extracted is the mechanism of action of a drug in a biological system and the dataset comprises two or more variables measuring the activity of the drug in said biological system. 
     
     
         13 . A method as in  claim 1 , wherein the information to be extracted is the identity of one or more biomarkers in a biological system and the dataset comprises two or more variables measuring the activity of a drug in the biological system. 
     
     
         14 . A method as in  claim 1 , wherein the information to be extracted is the one or more pathways that connect the drug to the one or more second variables through the one or more first variables. 
     
     
         15 . A method as in  claim 1 , wherein the dataset has been taken from measurements of the activity of a biological system. 
     
     
         16 . A method as in  claim 15 , wherein the biological system is a cell line, an animal, or a human. 
     
     
         17 . A method as in  claim 1 , wherein the information to be extracted is the mechanism of toxicity of a drug in a biological system and the dataset comprises two or more variables measuring the activity of the drug in said biological system. 
     
     
         18 . A method as in  claim 1 , wherein data reflecting the use of two or more drugs in the same biological system are included in the dataset and wherein the information to be extracted is the mechanism of action of the two or more drugs working together in said biological system. 
     
     
         19 . A method as in  claim 1 , wherein data reflecting the use of two or more drugs in the same biological system are included in the dataset and wherein the information to be extracted is the mechanism of toxicity of the two or more drugs when used together in said biological system. 
     
     
         20 . A method as in  claim 1 , wherein data reflecting the use of two or more drugs in the same biological system are included in the dataset, the dataset comprises two or more variables measuring the activity of the drug in the biological system, and wherein the information to be extracted is the identity of one or more biomarkers of the two or more drugs' efficacy together in a biological system.