Each well on the plate holds one gene variant of the starting
protein, expressed and screened for activity (fitness). Error-prone
PCR introduces random point mutations into the gene each round —
most are neutral or harmful, but a few improve function. DNA
shuffling instead recombines fragments from several already-good
parents, converging faster but exploring less new sequence space.
After screening, only the fittest fraction of wells survive to
seed the next round's library — an iterated cycle of diversify →
screen → select, exactly like natural selection compressed into
days instead of millennia.
fitness(child) = fitness(parent) × (1 + N(0, mutation_rate))
survivors = top (1 − selection_pressure) fraction by fitness
diversity(round) = std(fitness) / mean(fitness)
- Mutation rate — how much each new variant's fitness can drift from its parent; too high mostly breaks the protein, too low finds nothing new.
- Selection pressure — fraction of the library discarded each round; higher pressure converges faster but risks losing diversity needed for later improvements.
- Method — error-prone PCR explores broadly from one parent; DNA shuffling recombines multiple good parents for faster, more targeted gains.
The double helix above the plate lights up new mutation sites
each round — a reminder that every colour change on the plate
traces back to a real change in DNA sequence.