NITMB-SkAI New Faculty and Fellow Open House

Hastings Abstract:
The question of how species coexist is one of the most fundamental questions in ecology. The standard approach for using models to understand what leads to ecological coexistence can be caricatured as start with a model, specify parameters, and look at the outcome.  This can also be extended to looking at ranges of parameters.  If simulation approaches are used, an issue is the extent to which the results are sensitive to parameter choice.  Much recent work has used model fitting approaches to match a model to a particular outcome.  The model fitting approach has the feature that it explains a particular case, but this approach does not lead to general understanding.  I will use a different strategy based on an inverse approach of specifying a set of outcomes and then determining the range of parameters explaining potential observed behavior.  I will then apply these ideas to three specific cases; 1) the maintenance of large diverse food webs, 2) higher order interactions as an explanation for coexistence, 3) competition-coexistence tradeoffs as a mechanism leading to coexistence.  I will present possible further extensions.

Weinstein Abstract:
Massive improvements in DNA sequencing capacity over the past two decades have forced the field of molecular biology to continually reassess the role DNA itself ought to play in biological measurement. DNA and RNA are, on the one hand, biological machinery and, on the other, a vital laboratory medium for amplifying signal. We discuss one particularly powerful instance of these two roles coming together, showing how DNA can be engineered to encode large physical networks of molecules in biological tissue. Samples are first seeded with random, unique DNA sequences. A chemical reaction then pairs each sequence with its neighbors, so that the sequences self-assemble into a vast intermolecular network that encodes the geometry of the sample around it. A sufficiently dense map of these contacts allows relative positions to be reconstructed. We demonstrate this in whole zebrafish embryos and show that unsupervised likelihood maximization decodes the three-dimensional geometry of more than 1e7 molecules in individual specimens, down to length scales of 1-2 microns. Turning to the general problem of decoding geometry from this class of molecular networks, we close by surveying open problems at the interface between the engineering and the theory of reconstruction.