Sequestration-based Protein Neural Networks Tolerate the Effects of Shared Translational Resources

In bacterial protein neural networks built with molecular sequestration, ribosome competition rescales effective weights uniformly when mRNA ribosome-binding affinities match, preserving decision-boundary geometry.

• Why it matters: Shared translational resources can reshape biochemical computation, but the rules linking ribosome competition to classifier behavior are not well quantified.
• What they did: The authors develop a resource-aware mathematical framework for classification networks in bacterial systems, analyzing homogeneous versus heterogeneous ribosome-binding affinities and extending numerically from a single perceptron to multi-layer models with two hidden-layer nodes.
• The result: With differing affinities, competition induces predictable decision-boundary curvature confined to an input region set by binding-affinity variability, and in multi-layer networks curvature distortions propagate within a bounded uncertainty region.

AI summary of the abstract · may contain errors · not medical advice · read the paper.