Biomolecular condensates and the “binding problem”

Cells acquire diverse sensory data about themselves and the complex environments in which they act. If a sensory apparatus can distinguish only a small number of qualitatively distinct internal and external states, it seems best to hard-wire inputs to behavioral outputs. However, an escalation in the number of distinctions a system is capable of making eventually results in a new challenge for the system: to make sense of its sensory inputs in a context-dependent fashion. This project aims at constructing a biochemically and biophysically informed computational model to study how this might happen. The project is guided by an analogy and a hypothesis. The guiding analogy asserts that the task faced by a cell is a small-scale analog of what cognitive neuroscience calls the "binding problem", i.e. the linking together of features that are correlated by the process of observation. A crucial ingredient of hypothetical solutions to the binding problem is an attention mechanism—a mechanism that "highlights". The hypothesis asserts that a potential attention mechanism for feature-binding results from the empirically observed phenomenon of liquid-liquid phase separation, which creates transient membraneless compartments ("droplets") of scaffolding proteins that might regulate signaling cross-talk. A specific objective of modeling is to appreciate the forms of regulation, induction, amplification, suppression, and memorization of cross-talk a droplet might enable. The challenge is to proceed at a level of abstraction that is grounded yet not suffocated by unnecessary detail and hence computationally explorable for the benefit of conceptual insight. This is achieved by leveraging and adjusting a modeling language based on graph-rewriting that we developed over the past decade. A minimalist experimental sub-aim will assess the basic soundness of the modeling approach.