Funded Grants

Origins of innovation in tinkered networks

When the explorers reached the end of the known World, in those early times of exploration, they often went back talking about monsters. In those days, there were empty spaces in the maps, labeled as “Terra Incognita”, and the monsters inhabited those blank spaces. As explorations advanced, the maps became more precise, our knowledge better and the monsters vanished. So far, we have no news of their return, and we only see them in old engravings, movies and in our imagination. Somehow, not everything seems possible. However, nature abounds in complex forms and structures and seems to have an infinite power of generating complexity. We are fascinated by their richness and diversity and wonder how such complex entities have evolved. And yet, biological complexity is the result of an apparently inefficient mechanism of change: tinkering. Indeed, evolution operates by extensively re-using previous structures, and it is unable to foresee the future, as an engineer would. For example, proteins used to build the lenses in our eyes are also used somewhere else as enzymes. Genes involved in triggering the death of cells during the formation of fingers in vertebrates are later recruited to participate in the formation of the nervous system. Moreover, engineers can invent a completely new device with no inspiration from previous designs. These differences were pointed out in 1976 by the french biologists François Jacob. His paper received great attention, and his conjecture of understanding evolution as a tinkerer received great support over the last decades, particularly from developmental biology. But the reasons for the great success of tinkering remain essentially unknown. An additional feature of evolution is the presence of widespread convergence: common solutions are found to common problems: Evolution often re-invents similar structures and functional traits by tinkering from available components, as if only some special solutions could be achieved. Such convergent behavior of evolutionary innovations suggests that common rules shape the emergent of complexity in nature, imposing strong limitations to what is possible. What are those?

This project involves an exploration of the question of how tinkered evolution generates successful innovations and why these innovations usually converge to common solutions. In order to quantitatively explore this problem, we consider different levels of complexity, from cellular to ecological. All these systems are studied by considering them as a set of many interacting components forming a web. Such web, also known as a complex network, has been shown to display common patterns of organization, often resulting from simple rules of duplication and rewiring. Here by duplication we mean that random copies of already present existing are generated. This can occur through mutations and the resulting redundant copy can afterwards acquire new functions. A given protein for example, is linked to a number of other proteins by physically interacting with them (at some point in time and space). If a mutation occurs such that the gene coding for the protein becomes copied twice and redundant information will be present. But if mutations in the gene occur, they can rapidly modify the repertoire of interactions of the redundant copies and new functions can emerge. In this way, tinkering provides a source of innovation. In a different context, complex ecosystems will change over time by “inventing” new species: here speciation is the mechanism introducing copy, whereas further coevolution introduces changes in the redundant species (through diet specialization, for example). As a result of copy-and-change dynamics, complex networks are generated. The resulting networks display strong similarities, including the presence of small world properties (namely short chains connecting any pair of elements) and heterogeneity: most elements have a few links whereas a handful of them (the hubs) have many connections. Together with these structural properties, tinkered networks (as we call them) exhibit a high fragility under the removal or damage of hubs and a high robustness under random mutations. The presence of these common traits, often measurable in quatitative terms, might pervade the convergent designs found in nature.

Understanding the origins of this dual character of tinkering, that is to say its efficiency and limited repertoire, can be achieved by studying the underlying landscapes where these evolutionary paths take place. Fitness landscapes have been extensively used as a powerful metaphor of evolutionary optimization. A landscape has been often visualized as sets of connected valleys and mountains, whose height represent the fitness of a given combination of biologically relevant traits. Evolution would take place by exploring the universe of possible solutions and discarding those having smaller fitness through some selection process. Combinations of traits on the valleys would be rapidly eliminated, whereas those close or on top of mountain peaks would be favoured. Not much is known about the real nature of these landscapes in the real world, and only for RNA molecules a good understanding has been achieved. In this case, the linear RNA sequence, the genotype, folds in space into a given shape. Such shape is the phenotype, which is the link with functionality: a given folded RNA is associated to a given activity. These landscapes have been found to be neutral, meaning that many single mutations do not change the phenotype, and thus leading to neutral changes. Populations in this case spread through landscapes having “flat” domains where no change in fitness is found, followed by rapid changes when some fittest solution is found at the boundaries of the flat region. Our recent work using networks performing computations has shown that such landscapes might be the rule in totally different contexts. Perhaps such universal features pervade the presence of convergent designs.

Our project will study the nature of these landscapes associated to tinkered networks, at the cellular, tissue, developmental and ecological levels. They are all connected: cellular networks are responsible for cell differentiation and thus cellular diversity, which is required to generate and sustain efficient tissues. Cell diversity and its spatial organization in tissues are at the heart of development, and we want to see what types of networks allow reliable and flexible development to occur. Finally, an appropriate comparison with the real world needs considering what ecologist Evelyn Hutchinson dubbed “the ecological theater”: phenotypes make sense only if performing their role in a well-defined physical context. A model of development of embodied organisms will be constructed in order to explore the requirements for pattern formation to allow animal diversity to emerge and flourish, and determine the role of emergent dynamics versus selection under tinkered evolution. Such a multi-level approach to evolution will help answering some of the key fundamental questions of evolutionary ecology and, perhaps, why the monsters are gone.