New preprint: When a new node (novelty) is added to a (protein) network, how (and why!) does that network's resilience change?
Work from the @sfiscience CSSS, led by @aprilkleppe w/ @lholmerx @kesm_ith @mmjohnEEB @anshuman2111 @laura_stolp @Ashley_Teufel
https://www.biorxiv.org/content/10.1101/2020.07.02.184325v1
Work from the @sfiscience CSSS, led by @aprilkleppe w/ @lholmerx @kesm_ith @mmjohnEEB @anshuman2111 @laura_stolp @Ashley_Teufel
https://www.biorxiv.org/content/10.1101/2020.07.02.184325v1
Some deets:
We got the idea of network "resilience" from a great paper ( https://www.pnas.org/content/116/10/4426) by @marinkazitnik and co.
It's basically the entropy of the component sizes distrib as you iteratively disconnect more nodes
(sidenote this certainly could be called robustness) 1/
We got the idea of network "resilience" from a great paper ( https://www.pnas.org/content/116/10/4426) by @marinkazitnik and co.
It's basically the entropy of the component sizes distrib as you iteratively disconnect more nodes
(sidenote this certainly could be called robustness) 1/
Their paper showed that different species' protein networks can have a range of resilience values, which gave us an idea:
If resilience is defined by removing nodes, how does the resilience of protein networks change when adding new nodes?
A
presilience
if you will 2/
If resilience is defined by removing nodes, how does the resilience of protein networks change when adding new nodes?
A


We're imagining evolutionary novelty as the introduction of something new into a system, which doesn't in turn destroy the organism. *Of course, it's not this simple.
But in our context, we're curious about the node attachment mechanisms that increase a network's resilience. 3/
But in our context, we're curious about the node attachment mechanisms that increase a network's resilience. 3/
So! We simulate a bunch of new nodes being added to a few different ribosomal protein networks under three different node-attachment mechanisms
- uniformly at random
- degree-based preferential attachment
- *gene-expression-based* pref. attach.
/4
- uniformly at random
- degree-based preferential attachment
- *gene-expression-based* pref. attach.
/4
I found this interesting: the presilience is highest when new nodes (proteins) preferentially attach to nodes already in the network *proportionally based on the nodes' gene expression*
This suggests a potential relationship between gene expression & protein network structure /5
This suggests a potential relationship between gene expression & protein network structure /5
Then there's a whole fun SI section where we look at the resilience of different toy networks as we vary their parameters.
We also show that the maximum value for a network's resilience is 0.5 (previously described as 1.0)
/6
We also show that the maximum value for a network's resilience is 0.5 (previously described as 1.0)
/6
And some open code! https://github.com/jkbren/presilience
In the repo, we actually go into a little more detail and toy examples in three tutorial-esque notebooks.
Also included the matplotlib code for making the figs. /7
In the repo, we actually go into a little more detail and toy examples in three tutorial-esque notebooks.
Also included the matplotlib code for making the figs. /7
tl;dr - here's a pic that the 8 of us took in Santa Fe.
Getting to do science with your buddies is one of the best parts of being a scientist.
8/8
Getting to do science with your buddies is one of the best parts of being a scientist.
8/8