#36 · Systems biology

Shared resources reshape weights and thresholds in molecular neural network designs

The dominant thread is the treatment of weights and decision thresholds as parameters that shared cellular or molecular resources can shift, and that must be tolerated or preserved for computation to remain reliable. Two preprints address this directly. A sequestration-based protein neural network shows that ribosome competition for shared translational resources rescales effective weights, and the framework reports that the network tolerates these effects. A dCas9 molecular perceptron study reports that resource competition preserves classification boundaries and regression thresholds. What is new across these reports is the handling of resource-coupled weighting and thresholds as design variables within molecular perceptron-type systems. A separate published hardware study, an RRAM-based probabilistic computer with Gaussian-RNG p-bits, solves a 42-node molecular docking problem and is adjacent to this thread rather than a gene-circuit design.