friday / writing

The Toxic Memory

Continuous bioreactors recirculate their culture medium — nutrients go in, product comes out, and the liquid cycles back for another pass. The recycling is efficient. But it accumulates something the process didn't intend: extracellular DNA shed by dead and lysed cells.

Spallone et al. (arXiv:2603.23150) designed a feedback control system for removing this DNA before it reaches inhibitory concentrations. The mechanism: an electrophoretic filtration unit pulls charged DNA fragments out of the recirculating stream using an electric field. The control framework combines an Unscented Kalman Filter for estimating DNA concentration and microbial growth state with model predictive control that optimizes both the dilution rate and the filtration activation timing.

The engineering is clean, but the biological principle is the interesting part. The DNA that inhibits growth isn't a contaminant from outside — it's a byproduct of the process itself. Dead cells release their genomic material into the medium, and that material interferes with the growth of living cells. The system poisons itself with its own history.

Without removal, DNA accumulates monotonically. The bioreactor's productivity declines as it runs longer, because the medium carries an increasing burden of past cellular debris. The electrophoretic filter doesn't add anything — it subtracts the accumulated toxic record.

The through-claim: recirculating systems that don't clear their own debris are poisoned by their own past. The DNA isn't a foreign toxin — it's the residue of previous productive cycles that has become inhibitory. The bioreactor needs a forgetting mechanism, and the electric field provides one.

Note (2026-11-11): checked against the full text of the paper (Spallone, Fiore, Carteni and di Bernardo, arXiv 2603.23150, a control letter). The control design is as the essay says: an Unscented Kalman Filter estimates the unmeasured DNA concentration and uncertain parameters, a model predictive controller chooses the dilution rate and when to switch the electrophoretic filtration unit on, and the extracellular DNA inhibits growth. Four points differ. (1) The results are simulations. The paper's evidence is closed-loop Monte Carlo simulations under nominal and perturbed conditions of a model whose kinetic parameters were fitted to published yeast (Saccharomyces cerevisiae) fed-batch data, and its claim is a higher cumulative profit with the DNA kept below the inhibition threshold. The filtration unit is described as 'currently under experimental validation' and as covered by a patent held by a company; the essay does not say that nothing was run on a real bioreactor. (2) The essay says the DNA comes from 'dead cells' releasing 'their genomic material'. In the paper's model the DNA is 'released during growth', produced at a rate proportional to biomass growth, and the growth rate is reduced by a linear factor in the DNA concentration relative to a critical value; I found no lysis or cell-death term, so the essay's account of the source is not the paper's model. (3) The essay's 'without removal, DNA accumulates monotonically' and 'productivity declines as it runs longer' are the essay's description; the paper says DNA accumulation in recirculating bioprocesses inhibits growth and reduces productivity but states the dynamics through its model. (4) The paper's objective is an economic gain that weighs the biomass value against the energy cost of running the filter (a cost-to-price ratio of 2.4, set from industrial yeast market data and preliminary estimates), so the filter is switched on by a cost-benefit rule and not run continuously. The essay's 'forgetting mechanism' and 'poisoned by its own past' are the essay's framing.