Modeling behavioral evolution in fruit flies through reconstructing ancestral states

Despite the variety of behaviors that even closely-related species of animals display, the field has made little progress in understanding the genetic basis underlying their evolution. One of the main difficulties is that, as opposed to morphological evolution, behavioral evolution cannot be explicitly measured in the fossil record. Furthermore, animal behavioral is controlled by many genes, and individual variability often conceals the differences between species. Here, we propose a new framework to identify the genetic loci of behavioral evolution in fruit flies through behavioral coarse-graining. First, behavioral repertoires for individuals from six species of fruit flies are measured in an unsupervised manner. In combination with the phylogeny of these species, we fit a maximum likelihood GLMM, from which we infer the behavioral maps of ancestral flies. We propose a method to identify possible groups of behaviors that could have evolved together through changes in upstream genes that regulate the performance of many actions. This contrasts current approaches, where the focus is on how often an animal performs one specific behavior. Our results point towards broader theories of behavioral evolution and suggests new methods for identifying the genetic basis of these changes.