Eduardo D. Sontag's latest preprint, posted on arXiv, makes a case for shifting systems biology away from steady-state thinking. The paper argues that overshoots, biphasic dynamics, adaptation kinetics, fold-change detection, and cumulative exposure effects are often what determine phenotypic outcomes—yet standard dose-response analyses miss them. Sontag frames these as "dynamic phenotypes" and treats them as input-output properties that constrain underlying network architectures independently of precise parameter values.
The core theoretical move is to use control-theoretic concepts—monotonicity, equivariance, and input-output analysis—as precise mathematical tools rather than loose metaphors. Monotone systems, for instance, yield sharp impossibility results: if a network is monotone, certain transient behaviors cannot occur, so observing those behaviors falsifies entire model classes. In contrast, architectures with antagonistic pathways, particularly incoherent feedforward (IFF) motifs, naturally generate non-monotonic and adaptive responses. Sontag shows these motifs recur across molecular signaling, immune regulation, and population dynamics.
Beyond step inputs, the paper highlights experimental probes that are often underused: periodic forcing, ramps, and integral-type readouts like cumulative dose responses. These can separate competing motifs, expose invariances such as fold-change detection, and reveal structure that steady-state experiments hide. The overall argument is a call to reorient the field—from asymptotic behavior toward transient, input-driven dynamics as the primary lens for understanding, testing, and reverse-engineering biological networks. The paper is available as arXiv:2512.24945.{