String network additional interactions compared to network expansion

Question

[Irene Hemel]

I have a list of proteins form which I want to create a ppi network with additional proteins using the StringApp in Cytoscape. I figured out that there are two different ways to do this. The first way is to make a string network with the input proteins, which can then be expanded via Apps -> STRING -> expand network. The second way is by specifying a maximum number of additional interactors when creating the initial network.

These two methods create networks containing different additional proteins and large differences in the amount of interactions, with less interactions in the networks created by expanding it. Can anyone tell me what the difference is in the methods/algorithms behind the addition of these extra proteins, resulting in these different outcomes?

Answer 1

[scooter]

Hi Irene, Sure. In the case of loading additional proteins initially, you are just adding nodes that are connected to any of your proteins at a score better than or equal to the confidence cutoff. That much seems clear. In the case of expanding your network, there is an additional parameter -- the edge selectivity (or alpha value) that also gets factored in. The confidence score still gets used, but so does the alpha value, so you are likely to choose different nodes based on the combination of those two. I'm not sure exactly how the database processes the alpha value, to be honest, but by experimenting, the higher the value, the fewer (i.e. more selective) the edges are and the lower the value, the more edges you'll get. Of course, that means that you'll also get different nodes back as a result. Roughly speaking, a selectivity value of 0 should get you the same result as if you had asked for 10 additional interactions initially.

-- scooter