Kanda Euatham1, Natee Tongsiri
Reconstructing the regulatory relationships between genes using multiple time point expression profile data (EPD) is a powerful computational method to gain insight into gene networks. One such method uses binary on/off relationships to characterize the under- and over-expression of genes acting in unison. This approach uses only the relative expression levels of the genes of interest at multiple time points. One aspect of the EPD these methods often fail to account for is the inherent variability in the measurements of the gene expression levels. We characterize the variability in expression levels for a single time point to measure the inherent variability in that dataset. We then generate multiple new expression profile data samples from the original data and measured variability. These new datasets are then binarized to test whether the gene network relationships change due to the random sampling. This also allows us to test different variation magnitudes to set limits on how large the inherent variability should be to yield reproducible results for the binary gene network method. We find that the current variabilities in EPD are too large to yield reproducible gene regulatory networks, but that the data for some particular genes are sufficient to generate reproducible binarizations.
Share this article