The models presented here could be useful in understanding development, as well as cell-fate reprogramming (which can be induced artificially, but has also been shown to happen naturally, Weimann et al., 2003). We have derived general results about the dynamics and co-expression properties of switch networks, and shown the flexibility of bHLH dimerisation networks. Of the networks studied here, which were generically formulated with usual kinetic equations, only a subset can co-express antagonistic elements at a similar level, higher than the basal level: those with mutual inhibition, autocatalysis, and leak (but only when the cooperativity is very close to 1, and the transcription strength sufficiently low), and bHLH dimerisation networks (when the competition is sufficiently weak). This restricts the classes of models which can reproduce experimentally-observed co-expression of antagonistic factors, as well as showing how it can occur.
Strikingly, even though bHLH networks are the most apt to coexpression of antagonistic elements, the selection of one element requires a transient increase in competition, which is not what is thought to happen over a long time scale in the course of differentiation. Transient, hitherto-unexplained increases in competition have however been observed in a few cell lines upon differentiation, and could be a general phenomenon.
In order to model specific differentiation events, these networks would probably need to be extended to take into account combinatorial interactions, which could complicate their behaviour. The models would also gain from being extended to take into account non-symmetrical networks, in which some switch elements are stronger than others, and stochastic kinetics.