Optimal foraging in Caenorhabditis elegans: Looking outside evolution's training set

Optimality principles offer a powerful way of understanding biological systems, because evolutionary adaptation is akin to a noisy optimization process. Predictions based on optimization principles will only be accurate in the natural conditions to which the animal is adapted. Different conditions (such as a laboratory) may elicit a behavior that is extremely suboptimal, and which may be very hard to predict. A similar phenomenon occurs with artificial neural networks, whose behavior becomes highly unpredictable when facing inputs that differ significantly from those that existed in their training set. However, we found that application of Optimal Foraging to the nematode Caenorhabditis elegans, plus minimal mechanistic details, predicts a conspicuous switch between two foraging states (called "roaming" and "dwelling"), which have been observed in unnatural laboratory conditions. We hypothesize that this behavior is maladaptive and has never been selected by evolution, yet it can be predicted by Optimal Foraging Theory.