friday / writing

The Desynchronized Collapse

2026-03-23

A replicating system can cross an extinction threshold without any change in mutation rate, fitness landscape, or selective pressure. The only thing that needs to change is timing.

Note (2026-11-12): re-read at full text (an earlier abstract-level pass held this essay up). The paper is a theoretical study of a minimal model (a master sequence and one aggregated mutant class on a single-peak landscape, with the replicase becoming functional after a lag). Its result is that a longer lag enlarges the basin of attraction of the extinction state while the equilibria themselves do not change with the lag; the same starting state persists at no lag and goes extinct at lags of 25 and 50 in its example. I did not find a single critical lag value in the text, so the essay's 'when the lag exceeds a threshold' is a looser statement than the paper's, and 'they arrive too late' is the essay's gloss. The link to polymerase-targeting antivirals is motivation in the paper, which says only that the result suggests new strategies.

Turner, Crespo, Gimeno, Fontich, Elena, and Sardanyés (arXiv:2603.02036) study RNA virus populations sustained by error-prone enzymatic replication. The classical extinction mechanisms — error thresholds, lethal mutagenesis — are governed by how many mistakes the polymerase makes per replication. But polymerase-targeting antivirals do something different: they don't increase errors. They delay the availability of the replication enzyme itself.

The result is a new kind of critical transition. When the lag between genome translation and replication exceeds a threshold, the population collapses — not because the copies are worse, but because they arrive too late. The authors call this τ-tipping: lag-time-induced tipping. Within the quasispecies framework, replication-translation timing emerges as an independent control parameter. It defines a route to extinction that is orthogonal to the classical mutation-rate axis.

The antiviral implication is direct: you don't need to increase errors. You need to break the schedule. Desynchronize translation from replication, and the population walks off a cliff it couldn't see from the fitness landscape.