Project: Develop a Dynamic Model that Incorporates Text-mining to

Reconstruct Networks   (NIH 1R01GM079688-01)

 

PI: Christina Chan, Department of Chemical Engineering and Material Science, Michigan State University

Co-PI: Rong Jin, Department of CSE, Michigan State University

 

Abstract:

Improved understanding of the mechanisms of diseases (e.g., lipotoxicity or steatosis) and the identification of effective drug targets require better understanding of how diseases alter cellular processes from the healthy state. With advances in high throughput technology, profiles of gene expression, proteins and metabolites can be acquired to help elucidate the network of pathways involved in producing a specific phenotype. Many recent analyses attempt to infer causal relationships from high-throughput data. However, despite substantial advances and progress in our understanding of biological processes, genomics alone so far has failed to enhance the drug discovery pipeline. To address this, we have developed an integrative framework that reconstructs networks of active pathways from gene expression and phenotypic profiles. Although studies illustrate that the framework is able to predict those pathways that should be modified to reduce hepatic toxicity, transients, such as cycles and feedback loops, may not be well captured with the current framework. To address this shortcoming we are proposing to develop a knowledge-base, dynamic framework which should improve the gene selection process, and in turn improve the predictive capability of the framework. The eventual goal is to apply this framework to identify the pathways (potential targets) that may be altered to reduce the accumulation of intracellular lipids, e.g. steatosis, and inflammation in livers of rats that are fed a high fat diet.

 

Students

  1. Yang Zhou

 

Project Goal:

The eventual goal is to apply this framework to identify the pathways (potential targets) that may be altered to reduce the accumulation of intracellular lipids, e.g. steatosis, and inflammation in livers of rats that are fed a high fat diet. Specific aims are: (a) develop a novel approach that incorporates domain knowledge retrieved from the free text as well as gene expression data to predict cellular or phenotypic response; (b) develop an optimized dynamic Bayesian Network to infer gene regulatory networks from time series data; (c) experimentally validate the model predictions; and (d) characterize the livers from rats fed high fat vs. normal diets.

 

Current Results:

  1. Develop a Bayesian framework for gene selection that explicitly take into account the noise and uncertainty in gene expression data

  2. Develop knowledge driven matrix factorization framework for network reconstruction

 

Publication:

 

Journal:

  1. X. Yang, Y. Zhou, R. Jin, C. Chan, Reconstruct Modular Phenotype-specific Gene Networks by Knowledge-Driven Matrix Factorization. Bioinformatics 25(17): 2236-2243, 2009
  2. S. Srivastava1, L. Zhang, R. Jin, C, Chan, A Novel Method Incorporating Gene Ontology Information for Unsupervised Clustering and Feature Selection, PLoS ONE, 3(12):e3860, 2008
  3. R. Jin, L. Si, and C. Chan, A Bayesian Framework for Knowledge Driven Regression Model in Micro-array Data Analysis, to appear in International Journal of Data Mining and Bioinformatics (IJDMB), 2(3): 250-267 2008

Conference:

  1. Y. Zhou, Z. Li, X. Yang, L. Zhang, S. Srivastava, R. Jin and C. Chan, Using Knowledge Driven Matrix Factorization to Reconstruct Modular Gene Regulatory Network, Proceedings of 23rd National Conference on Artificial Intelligence (AAAI 2008), pages 811-816, 2008

  2. R. Jin, L. Si, S. Srivastava, Z. Li, C. Chan, A Knowledge Driven Regression Model for Gene Expression and Microarray Analysis, Proceedings of the 28th IEEE Engineering in Medicine and Biology Society (EMBC 2006), 2006