I am a quantitative biologist with a background in Physics who is broadly interested in discovering the fundamental principles that govern living systems. Echoing Phil Anderson, More is Different: Biology has a beautiful wealth of complexity that operates at many different scales with interesting structures underneath. I am currently lucky to sit at the KITP doing science by the beach.
My research at the KITP has focused primarily on one project: how do cells in a developing embryo encode information about where they are in space? The current paradigm of positional information is the "french-flag" model, in which a single diffusible molecule is sourced at one end of the embryo and is degraded throughout. At steady state, the concentration of this molecule (often called a morphogen) exponentially decays from the source and thus provides a simple readout of spatial position.
Using advances in scRNAseq technology, I am working to understand this problem from a new direction: what if the state of cellular gene expression within a developing embryo changed collectively as a function of position. When viewed through this lens, positional information becomes akin to Manifold Learning with the expectation that scRNAseq data should lie on a low-dimensional manifold parameterized where each cell was sampled from the embryo.
In the past, I was interested in understanding how horizontal gene transfer and homologous recombination quantitatively affect evolution, with the ultimate goal in forming simple models able to predict rates of evolutionary outcomes. Population genetic models exist at both extrema of recombination rate; the field has some quantitative grip on the fully linked, asexual evolution and fully decoupled, single locus evolution. We currently have no satisfactory models, and thus computational tools, in the intermediate regime, let alone that account for the structural variation observed in natural samples. Due to their smallish genome sizes, short (on the time-scale of a research project) generation times and epidemiological relevance, these questions are best addressed within the microbial world.
With the recent introduction of long-read, single molecule sequencing this is no longer a purely theoretical pursuit but can be empirically measured within 'wild' or 'clinical' ecosystems. My past research attacked this problem using both empirics and theory driven modeling; I find one informs the other quite well. Evolution is a noisy process and thus requires large datasets to find signal. This requires scalable computational algorithms. Useful heurestics can teach us a lot about the underlying structure of the problem. My contribution in this direction resulted in the algorithm PanGraph.
I completed my PhD in UCSB where I studied under Boris Shraiman in the tangentially related field of Morphogenesis. Developmental Biology has been revolutionized by the fluorescent molecule which allows researchers the ability to watch development happen live. However, the mechanical state of the embryo which ultimately drives the developmental flow remains unknown. Boris and I worked out a model of the active mechanics of cells, and leveraged this to formulate a robust mechanical inference algorithm solely from measured cellular geometries obtained from live image movies, that was shown to be quite predictive during early Drosophila gastrulation. We also found a pretty neat duality.