Looking at stochasticity in nodes and in functions within GR networks. For T-helper differentiation, the GR network is reasonably complex (he’s using this as the example for the talk). How does the system behave when it is subjected to stochastic behaviour? Start by looking at the robustness of the system. Robustness is maintenance of specific functionalities of a system against perturbations.
Stochastic modelling techniques include static network analysis, kinetic modeling techniques, and boolean modeling (randomly flipping the expression of nodes). It is the latter technique that is used in the talk. If you’re given a small GRN, each gene is either active or inactive, e.g. B is activated by A. And/or nodes can be represented by these boolean functions as well.
In their previous work, they looked at stochasticity in nodes (SIN). With a node, you can have a stochastic transition out of the current state to another state, and we can go from SS2 to SS3 (for example) with some probability that can be calculated. Single fault model mentioned. When the system is subjected to stochastic behaviour, there can be multiple transition choices with various probability. With SIN, they got stochastic cellular differentiation. You understand what the probabilities are by looking at a population of cells.
The SIN method overrepresents noise in the system, therefore they started working on SIF (stochasticity in functions), which is what this paper describes. The noise depends on the complexity of the biological function and also on the input nodes. The SIF model specifies biologically meaningful transition rules. With SIF, you can calculate the probability of the fault in a function (depends on the complexity of the function).
Measure of robustness: probability of a steady state returning back to itself in the presence of internal pertubations (or faults). Once you start introducing faults, the robustness is much lower the more faults you add for SIN, but the robustness is only marginally lower using SIF.
You can then infer robust GRNs using ranking and SIF. They use full exploration of the state space instead of sampling. Implicity methods are based on Binary Decision Diagrams. Use the GenYsis toolbox.
Please note that this post is merely my notes on the presentation. They are not guaranteed to be correct, and unless explicitly stated are not my opinions. They do not reflect the opinions of my employers. Any errors you can happily assume to be mine and no-one else’s. I’m happy to correct any errors you may spot – just let me know!