Abstract: Transcriptional regulatory network (TRN) discovery using information
from a single source does not seem feasible due to lack of sufficient
information, resulting in the construction of spurious or incomplete TRNs. A
methodology, TRND, that integrates a preliminary TRN, gene expression data and
gene ontology is developed to discover TRNs. The method is applied to a
comprehensive set of expression data on B cell and a preliminary TRN that
included 1,335 genes, 443 transcription factors (TFs) and 4032 gene/TF
interactions. Predictions were obtained for 443 TFs and 9,589 genes. 14,616 of
4,247,927 possible gene/TF interactions scored higher than the imposed
threshold. Results for three TFs, E2F-4, p130 and c-Myc, were examined in more
detail to assess the accuracy of the integrated methodology. Although the
training sets for E2F-4 and p130 were rather limited, the activities of these
two TFs were found to be highly correlated and a large set of coregulated genes
is predicted. These predictions were confirmed with published experimental
results not used in the training set. A similar test was run for the c-Myc TF
using the comprehensive resource www.myccancergene.org. In addition,
correlations between expression of genes that encode TFs and TF activities were
calculated and showed that the assumption of TF activity correlates with
encoding gene expression might be misleading. The constructed B cell TRN, and
scores for individual methodologies and the integrated approach are available
at systemsbiology.indiana.edu/trndresults.
Keywords: Transcriptional regulatory network, gene expression data analysis, B cell, transcription factors, gene ontology
