new food
how much better

inside the cell: raw material against ribosomes

biomass, as a share of the best possible

the control: ribosome share α(t)

The model, and Bellman's equation

p is raw material (precursors), r the protein machinery that makes proteins (ribosomes), both as shares of the cell. E is how good the food is. α, the one thing the cell controls, is the share of new protein it makes as ribosomes rather than as metabolic enzymes. Growth μ is ribosomes times how well fed they are. Time is in units of the ribosome's own rate.

A dot means the rate of change in time: ṗ = dp/dt, ṙ = dr/dt, given by the first equation above. V is the most extra biomass (in logs) still obtainable from where the cell stands. Because α enters the equation linearly, the best α is always a corner: all ribosomes where one more ribosome is worth something (∂V/∂r > 0), none where it is not. The amber curve is where ∂V/∂r = 0. Near the target the optimal cell switches back and forth across it: the chattering the control strip shows. The optimal policy is computed on a grid, so shares within about 0.2% of 100% are ties.

optimalsolves Bellman's equation on the whole plane and follows its policy
on–off (ppGpp)switches ribosome making fully on below the teal dashed curve g(p), fully off above it. E. coli's alarm molecule ppGpp works this way
precursor-onlysets α = g(p): the share that would be best at the current level of raw material
nutrient-onlyjumps at once to the best steady share for the new food, and stays there
no reactionkeeps the share that was best for the old food