Gene Network Reconstruction By Integration of Biological Prior Knowledge

Constructing gene association network from gene expression data is challenging, especially when the gene number is much higher than the sample size of the data. Assuming gene expression data are from a multivariate Gaussian distribution, graphical lasso algorithm is able to estimate sparse inverse covariance matrix by a lasso (L1) penalty (Friedman et al., 2007). The inverse covariance matrix can be seen as direct correlation between gene pairs in the gene association network. In our work, instead of using a single penalty, we apply different penalty values for gene pairs based on the priori belief whether the two genes should be connected or not. By incorporating prior knowledge, the weighted graphical lasso (wglasso) outperforms the original graphical lasso both on simulation studies and data from yeast and Arabidopsis. This approach can be applied to other plant and animal data sets where data dimensionality can be problematic.