Unraveling phenotypic variation in multiscale behavioral dynamics

Unraveling phenotypic variation in multiscale behavioral dynamics involves understanding the differences in internal and external parameters that lead to natural selection. The challenge lies in the nonlinear evolution of behavior across multiple timescales. The goal is to reconstruct phenotypic variation from limited behavioral observations without specifying the scales for comparison.

A new multiscale dissimilarity has been introduced to compare dynamics at all statistically resolvable resolutions. This is accomplished by creating reduced transfer operators at progressively finer scales and weighting their differences based on finite-sampling uncertainty. The dissimilarity defines a geometry of dynamical phenotypes, which is then reconstructed using diffusion maps.

The reconstructed phenotypic coordinates align locally with the parameter directions that dynamics are most sensitive to. Validation of the approach has been done using a stochastic model, followed by its application to bacterial and larval zebrafish behavior.

In bacteria, the study identifies run speed as a major factor in inter-individual variability, which is associated with the chemotaxis protein CheB. In zebrafish, the prior prey experience reshapes behavioral dynamics and phenotypic variability. Specifically, paramecia experience concentrates the fish in a region of phenotypic space related to prey capture. These findings provide a framework for identifying both internal and environmental variables that structure dynamical phenotypes across scales.