Thread by @yoginho

Just out in Biology & Philosophy: “Beyond networks: mechanism and process in evo-devo,” the first publication from a wonderfully productive collaboration with philosopher James DiFrisco.

researchgate.net/publication/33…

link.springer.com/article/10.100…

#philsci #evodevo

This paper should be of interest well beyond the field of #evodevo. Our argument applies equally to network-based explanations in systems biology. /1

The main point is that static representations of regulatory networks do *not* provide causal-mechanistic explanations for metabolic, physiological, developmental, behavioural, or other complex phenotypes. /2

To explain how a phenotype is generated, we must understand the underlying regulatory processes. This requires not only mechanistic decomposition of network structure, but also dynamical modelling to understand how network interactions synergise to produce the phenotype. /3

In particular, we identify three problems with static explanations in terms of gene regulatory networks:

(1) the problem of genetic determinism,

(2) the problem of correspondence,

(3) the problem of diachronicity.

/4

The problem of genetic determinism is simple: there is no set of genetic instructions that suffices to generate a particular phenotypes. It is hard to precisely specify what those instructions are. And: cellular and external context is crucial for the functioning of a network. /5

The problem of correspondence and the problem of diachronicity both have to do with the fact that networks unfold their effects over time, and that evolution at the network & phenotypic level are to some extend dissociable. /6

The problem of correspondence: it is impossible to precisely link phenotypic traits with components of an underlying network & their interactions. Trait definitions are question-dependent. Plus: conserved traits can be generated by divergent mechanisms due to network drift. /7

The problem of diachronicity: structure does not determine behaviour & therefore function of a network. Dynamics are radically dependent on context (boundary conditions). In order to understand a trait, we must explain how it is generated *over time* in its specific context. /8

The problems of determinism, correspondence, and diachronicity make it impossible to identify specific genetic network structures that represent a phenotypic trait. This is not a practical issue. The problem is impossible to solve *in principle*. /9

In order to overcome this limitation, we must move beyond explanations based on static network graphs to explicitly study the contextualised regulatory dynamics that generate phenotypic traits. /10