Zeros and Ones[draft]

We’re beginning with networks for two reasons. Gene regulation sits at a nexus between developmental and evolutionary time-scales. Which genes are turned on or off plays a crucial role in how multicellular organisms develop. But how they are turned on or off is both a complex, dynamic process that occurs over developmental time, and depends not only on other genes, but on the environment. We also know that changes to the genome affecting only regulation, rather than modifying proteins, is sufficient to induce significant change over evolutionary scales. A key insight was that how genes are put together, rather than what proteins are transcribed, determines the developmental and evolutionary outcomes. TODO: find a citation for this.

Importantly, for the task here, there is a long history of modeling gene regulation simply. A simple way to model these networks is to treat each gene as simply being on or off. Boolean Networks have been used to model gene regulation since the 1960s (Kauffman, 1969). These models have a long history.

Boolean Networks

TODO: Fix all this. Think of a cell state as a network of genes that can be on or off. Each gene’s state at one step is determined by a simple logical rule that looks at some other genes. Those on/off updates, repeated across all the genes, create a dynamical system, falling into recurring patterns or attractors. Kauffman suggested those attractors correspond to cell types, giving a way for order to emerge from complex regulatory interactions.

However, there are important differences between AND programs and Boolean Networks. Network and activation are conflated. Endogenous vs exogenous. In Boolean Networks, the update rules are applied to all genes at each time step.

, each consisting of one or more CRMs (the binding operations) combined by logical operators (the combiners) to produce an effect on a target gene (the assignment).

It’s form is meant to capture the binding and transcribed portions of DNA, and its function is to capture the causal relationships between genes and their effects on the organism.

• Policies from reinforcement learning?
• Stimulus-response mappings from behavior-based robotics?
• A control system from cybernetics?