Our study on bistability in the gene expression of honey bees is published

Posted October 2023 by Bryan Daniels

Finding critical transitions in gene expression time series

In development, cells in an embryo start out all having the same DNA and similar behavior. As the cells grow and divide, they eventually take on distinct identities. The fact that there exist separate identities ("cell types") and not just a continuum of possible behaviors can be explained by nonlinearities in the dynamics of gene regulation. The identities correspond to distinct jobs that the cells have in supporting the functions of the entire organism.

Similarly, in honey bee colonies, individual bees start their lives all performing the same behaviors within the nest. In the course of aging and in response to cues from the environment, each bee will eventually transition to a different bee "type" that goes outside the nest to find food. This transition is visible not just in the bees' behavior but at the level of genes expressed by each bee.

More generally, a transition from one type of behavior into two separate types connects with ideas of bifurcations in dynamical systems or phase transitions in statistical physics. Using mathematical tools from these fields, we can find the specific shape of the probability distributions expected near such transitions. This allows us to more easily search for such transitions in gene expression data.

In this project, we successfully found the transition from in-hive to foraging behavior in bees by looking for the transition from monostability to bistability in gene expression data. We could do this even with a relatively small amount of data: just 16 bees per timepoint. We also show that it beats other standard methods like Gaussian mixture models because it uses the shape of the transition known from Landau theory in statistical physics.

In the future, our method could be used as a precise and efficient way to search for critical transitions in high-dimensional data across many examples, from cell development to neuroscience to social systems.

Many thanks to my wonderful collaborators on this project: Rob Page, Ying Wang, and Gro Amdam!

Look for more details in our writeup in PLoS Computational Biology: https://doi.org/10.1371/journal.pcbi.1010704