Evolutionary tuning of activation/repression ratio for memory optimization in gene regulatory networks

Determine whether biological evolution has tuned the activation/repression ratio in gene regulatory networks to an optimal value that maximizes temporal memory capacity, as suggested by the observed non-monotonic performance peaking near a repression percentage of approximately 40–60% in the recurrent core of Escherichia coli.

Background

The paper analyzes how the balance between transcriptional activation and repression affects the performance of the E. coli gene regulatory network (GRN) under the reservoir computing framework. The recurrent core of the GRN exhibits a repression percentage of 41%.

By varying the repression percentage in randomized versions of the reservoir, the authors find that performance in memory-demanding benchmarks (critical memory capacity and the 10th-order NARMA task) varies non-monotonically with repression, achieving best performance around 40–60% repression, encompassing the biological value.

Based on these observations and analogies to excitatory/inhibitory balance in neuronal networks, the authors explicitly conjecture that evolution may have tuned GRNs to an activation/repression ratio that maximizes memory capacity.