Pleiotropy and its reduction by modularity

Ten Tusscher and Hogeweg (2011) set out the standard reasoning. Developmental gene regulatory networks are often said to be modular, with different functions performed by largely independent network parts. On this view modularity reduces pleiotropy. A part involved in one function can malfunction, or be tinkered with, without producing failure in other functions.¹ The authors accepted that the reasoning sounds appealing. They also argued that little research had explicitly tested how developmental network modularity, robustness and evolvability relate in the evolution of complex body plans.² In their view it was unclear how modular such networks are, how the modularity evolves and what it looks like.

Ten Tusscher and Hogeweg tested the question in evolved networks. To check whether non-functional and redundant parts hid an underlying modular architecture, they also computed modularity values for core networks derived from the evolved ones. Q values of around 0.3 were obtained for both SF and SS type networks.³ McGregor and colleagues (2012) also used the term core network for what remains after pruning. They pruned evolved chemical networks and found that typically only 5 reactions were needed to reach a fitness within 10% of the whole network.⁴ Rink (2013) used the term differently, for a possible ancient pluripotency core network shared by planarian neoblasts and ES/IPS cells.⁵

Levin (2022) argued that multiscale competency architecture reduces apparent pleiotropy. In his example, a mutation moves the mouth and also has a useful effect elsewhere. In a creature without such competency, selection never sees the useful effect, because the displaced mouth ruins fitness. If the mouth can move to its optimal spot, selection can evaluate the other consequence independently.⁶ Bongard and Levin (2023) made a parallel case. They held that the competency of the substrate in regulative development and remodeling can neutralize the lethal side effects of many mutations, which allows other possibly beneficial effects to be explored.⁷ Here the protection comes from repair by the parts, not from the independence of network modules.

Other authors treat pleiotropy as something to be reckoned with. Pio-Lopez, Hartl and Levin (2025) related their account of aging to antagonistic pleiotropy and hyperfunction theory, which state that what is good for development could ultimately cause aging.⁸ Rorot (2026) cited Wimsatt's warning that the localization heuristic risks ignoring pleiotropy, meaning the multiple functions of a single structure.⁹ How far modularity or competency removes pleiotropy in real organisms remains unsettled.