Greetings,
I'm assuming that you are using stringApp here, since you mention confidence values, if not, please give us more information about how you are getting your interactions. In any case, as far as "when to stop", it depends on what you're looking for. First, you mention interactions, so I'm assuming you are creating a physical interaction network, not a functional association network. Second, 0.4 is a pretty generous confidence interval -- it might be good to change your confidence score to something a bit tighter. Finally, when you say you continue to get more nodes and edges, are you limiting your additional nodes to those that interact with your viral proteins, or are you getting nodes and edges that connect to host nodes and edges, but not your viral proteins?
-- scooter