Natural Selection Fails to Optimize Mutation Rates for Long-Term Adaptation on Rugged Fitness Landscapes

By Jeff Clune, Dusan Misevic, Charles Ofria, Richard E. Lenski, Santiago F. Elena, Rafael Sanjuán. 2008

144 citations recorded by the source corpus.

Mutations are required for adaptation, yet most mutations with phenotypic effects are deleterious. As a consequence, the mutation rate that maximizes adaptation will be some intermediate value. This abstract summarizes a previous publication in which we used Avida, a well-studied artificial life platform, to investigate the ability of natural selection to adjust and optimize mutation rates. Our initial experiments occurred in a previously studied environment with a complex fitness landscape (Lenski et al. Nature, 423, 2003) where Avidians were rewarded for performing any of nine logic tasks. We assessed the optimal mutation rate by empirically determining which unchanging mutation rate produced the highest rate of adaptation. Then, we allowed mutation rates to evolve and we evaluated their proximity to the optimum. Although we chose conditions favorable for mutation rate optimization (asexual organisms not yet adapted to a new environment), the evolved rates were invariably far below the optimum across a wide range of experimental parameter settings (Fig. 1). We hypothesized that the reason mutation rates evolved to be suboptimal was the ruggedness of fitness landscapes. To test this hypothesis, we created a simplified 'counting ones' landscape without any fitness valleys and found that, in such conditions, populations evolved near-optimal mutation rates (Fig. 2, top row). In contrast, once moderate fitness valleys were added to this simple landscape, the ability of evolving populations to find the optimal mutation rate was lost (Fig. 2, bottom two rows). Additional experiments revealed that lowering the rate at which mutation rates evolved did not preclude the evolution of suboptimal mutation rates (see original manuscript). We conclude that rugged fitness landscapes can prevent the evolution of mutation rates that are optimal for long-term adaptation because of the short-term costs of traversing fitness valleys. This finding has important implications for evolutionary research in both biological and computational realms.

Subfields: Ecology, Evolution, Behavior and Systematics, Genetics, Sociology and Political Science

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecology, Evolution, Behavior and Systematics

In Genetics

In Sociology and Political Science

Citations in the corpus

Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules (2019, background): “These issues are resolved in Population Based Training (Jaderberg et al., 2017), which builds upon both evolutionary strategies (Clune et al., 2008) and random search (Bergstra & Bengio, 2012)”
Population Based Training of Neural Networks (2017, background): “…that employ self-adaptive hyperparameter tuning to modify how the genetic algorithm itself operates (B¨ack, 1998; Spears, 1995; Gloger, 2004; Clune et al., 2008) – in these works, the evolutionary hyperparameters were mutated at a slower rate than the parameters themselves, similar to how…”
The Surprising Creativity of Digital Evolution: A Collection of Anecdotes from the Evolutionary Computation and Artificial Life Research Communities (2018, neutral): “Yet both philosophical arguments [23, 26, 27, 82, 109] and high-profile publications [2, 16, 18 – 21, 52, 59, 60, 83, 85, 86, 92, 141, 146, 148] suggest that digital evolution can be a useful tool to aid and complement the study of biological evolution.”
Evolvability Search: Directly Selecting for Evolvability in order to Study and Produce It (2016, methodology): “To implement self-adaptation, the NEAT genome is extended with three pairs of mutation settings, which themselves mu-tate at ﬁxed rates (to escape inﬁnite regress) [4,15].”
Parameter Control in Evolutionary Algorithms: Trends and Challenges (2015, background): “This latter effect has also been observed in artificial life settings for digital organisms with discrete genotype spaces [34].”
Illuminating search spaces by mapping elites (2015, background): “…encourage diversity through a variety of different techniques, including increasing mutation rates when the rate of performance improvement stagnates 1,2,17–19 , explicitly selecting for genetic diversity 2,20 or behavioral diversity 21–25 , or changing the structure of the population 26 .”