The Eichler lab created a network of interacting proteins more likely to harbor de novo [mutations in autism probands: http://www.nature.com/nature/journal/v485/n7397/full/nature10989.html]. They then performed follow-up sequencing to confirm that these genes are more likely to be mutated. I did not see a reference for the follow-up study (it may be unpublished), but [this review: http://www.ncbi.nlm.nih.gov/pubmed/24387789] briefly describes the follow-up (page 100).

I am not sure that their follow-up discovered something 'new and important' but it certainly confirmed that their protein network contained many genes likely to have de novo mutations in sporadic autism populations.

Here are two such examples from our lab, where networks are being inferred from the data and then they are being used for further prediction followed by experimental follow-up:

1. Perturbation Biology: Inferring Signaling Networks in Cellular Systems
2. Drug Synergy Screen and Network Modeling in Dedifferentiated Liposarcoma Identifies CDK4 and IGF1R as Synergistic Drug Targets