Relating Feed-Forward Loop Crosstalk to Robust Information Transport Across Transcriptional Networks

Evolved biological network topologies may resist perturbances to allow for more robust information transport across larger networks in which their network motifs may play a complex role. Although the abundance of individual motifs correlate with some metrics of biological robustness, the extent to which redundant regulatory interactions affect motif connectivity and how this connectivity affects robustness is unknown. To address this problem, we applied machine learning based regression modeling to evaluate how feed-forward loops interlinked by crosstalk altered information transport across a network in terms of packets successfully routed over networks of noisy channels via NS-2 simulation. We developed 233 topological features which distinctly account for the opportunities in which two feed-forward loops may exhibit crosstalk. Random forest regression modeling was used to infer significant features from this modest configuration space. The coefficient of determination was used as a primary performance metric to rank features within our regression models. Although only a handful of features were highly ranked, we observed that, in particular, edge connected feed-forward loops correlated substantially with an improved chance for successful information transmission.