The limits of biological experimentation

Our longest-standing line of work plays around with a connection between biological organization, fundamental principles of high-dimensional statistics, and experimental design. Here’s the idea: Complex living systems tend to be highly organized. When we collect high-content, high-throughput data on these systems, this organization manifests in a substantial amount of statistical structure. Statistical theory and algorithms (mostly related to compressed sensing) allow us to exploit that structure through highly efficient “compressed” experiments with wacky designs, where each observed experimental outcome corresponds to a linear combination of conventional experimental outcomes.

Canalization of developmental trajectories

One of the most fascinating concepts in biology is the idea of canalization: dynamic processes, especially during development of multicellular organisms, are “canalized” to robustly follow the same trajectory and produce the same output, even in the face of environmental and genetic perturbation. We are interested in a deeper understanding of canalization: identifying canalized paths “hidden” in high-dimensional space, understanding how each path is triggered and regulated, and understanding how evolution “carves out” these canals in the first place.

At the moment, much of our work is focused on technical fundamentals, especially the reconstruction of cellular trajectories using single-cell “RNA velocity” measurements. We also have a growing interest in the early evolution of cellular differentiation in complex multicellular systems, the nature of genetic and phenotypic variability, and genetic assimilation.

The evolution of biological processes

Very much related to the topic above, we are interested in how a basic unit of organization, or biological process, emerges from selective pressures. A central aspect of this work is to formalize how selection acts through non-trivial genotype-to-phenotype maps to evolve a system in a particular direction, rather than focusing on how selection acts on individual genes.