Speaker
Description
Structural identifiability in chemical reaction networks is often studied for kinetic parameters, but much less is known about identifiability of the stoichiometric structure itself. We study this question for non-autocatalytic mass-action biochemical reaction networks with fixed rate constants. We ask when the stoichiometric matrix can be uniquely recovered (identifiable) from the induced dynamics.
Our results show that non-identifiability is not arbitrary. It is tied to specific structural features of the network. When several reactions share a common reactant complex, the stoichiometric structure across those reactions need not be uniquely recoverable from the dynamics, even under idealised observations. Yet this failure is highly structured. It is confined to the shared-reactant part of the network and shaped by precise relations among the associated rate constants. This leads to a clean separation between identifiable and non-identifiable regimes, and gives a sharper understanding of what mass-action dynamics can and cannot reveal about biochemical network structure.
