The total interaction score corresponds to the posterior odds ratio of interaction calculated using a naïve Bayes network as described in Scott and Barton 2007.
Four different features are considered in the prediction and can each output their own interaction score:
- Expression data,
- Orthology interaction relationships,
- Protein features ( protein domains, post translational modifications and localisation ),
- Network topology.
Intuitively, the total score measures how much more likely two proteins are to interact than not to interact, given the evidence considered.
A total interaction score of 2 means that the probability of interaction is twice as high as the probability of not interacting.