Ah, apologies, I thought you were interested in horizontal transfer. Honestly, now I'm not sure what your question is: the herpes virus' dormancy, or it's DNA insertion. The DNA insertion is already well documented, as is the dormancy, although I have no idea as to the mechanisms for said dormancy. Well, thanks for the rabbit hole!
Not OP, but I'm doing something similar with huginn [0]; I use it to scrape deals and sent POST requests to an Azure ml web service, which is output as RSS.
It's been a while since I played around with tensorflow, how difficult is it to deploy a model as a web service? I'm currently using azure and looking for other options (besides prediction.io).
I'm currently using a least-squares linear regression model trained with n-gram (unigrams) feature hashing, based on various social media metrics, with a heavy weight on sharing. It spits out a number between 1-50, with 25 being the most common.
Pretty much everything is automated with Huginn[0]: web scraping, data formatting, interaction with Azure ML[1], all of which is fed into a wordpress install using the pluginhunt[2] theme.