The abstract is read at the publisher; the summary is JClub's.
Engineered Escherichia coli implements a genetically encoded local learning rule that updates analogue weights stored in memregulons via sublethal kanamycin–driven negative learning.
- Why it matters: Training physical neural networks in the same substrate is hard because most platforms cannot store and update weights locally within the material that performs computation.
- What they did: The authors build memregulons where coupled plasmids encode an analogue weight as the relative abundance of P1 within a P1+P2 pool, then use environmental promoter activation plus a shared sublethal kanamycin signal to drive differential growth and weight change.
- The result: Measurements across single-strain trajectories and single-cell experiments show activity- and variance-dependent weight updates, while mixed co-cultures use one shared negative learning signal to selectively rewrite active memregulons, enabling externally routed supervised tic-tac-toe lessons.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.