Identifying details that matter: fruit fly development, genetic regulation, and microbial ecology

Abstract

The wealth and complexity of the known microscopic detail of biological processes and pathways make the search for universality particularly challenging and appealing for a physicist. This dissertation investigates several examples drawn from three different biological contexts.

First, I discuss the gene regulatory network responsible for segment patterning in the fruit fly. The fruit fly embryo is one of the best-studied examples of precision in biological processes. However, a novel technique I developed with my collaborators demonstrates that even in this system transcription is intrinsically noisy, as previously observed in bacteria. Using single-molecule-precision measurements of the transcriptional activity of four critical patterning genes, we exhibit universality of expression noise parameters and show how precision is recovered through spatiotemporal averaging. On a theoretical level, I demonstrate how these experimental findings help understand the multi-tier architecture of the patterning network: the diffusion-mediated non-locality of transcriptional response makes a cascade of readouts the optimal gradient response strategy, even if each readout is intrinsically noisy.

Second, I investigate the importance of microscopic parameters of networks at the scale of their global function. The fields of neural and genetic networks have exactly opposite assumptions on the matter, the former concentrating on synapse strength and the latter solely on network topology. I present a class of simple perceptron-based Boolean models within which the relative importance of topology vs. interaction strengths becomes a well-posed problem. I show that optimizing interaction strengths is a better strategy of achieving high complexity, defined as the number of fixed points the network can accommodate, and comment on the implications for real networks and their evolution.

Third, I discuss the so-called 16S tag sequencing method of studying microbial communities. The standard approach to 16S data, which relies on clustering reads by sequence similarity into Operational Taxonomic Units (OTUs), underexploits the accuracy of modern sequencing technology. I present a novel, clustering-free approach that exploits cross-sample comparisons to achieve sub-OTU resolution, and demonstrate that this new level of detail can provide new insight into factors shaping community assembly.

Finally, I discuss some common themes in the conclusions from these projects.