Finally, we examined the interconnectivity of the top ranked subnetworks (based on MI-links), and how the network structure is related to associations with disease.

We then identified associations between these subnetworks and 6 different MI-linked traits. We did this by considering at the overall expression of all proteins in the subnetworks (as eigenvectors) and by considering the expression of the network regulators alone.

In this project, we used cis-pQTLs as instrumental variables to reconstruct a causal protein network from measurements for 7532 serum proteins, from 5376 Icelanders. The global network was composed of 185 subnetworks, each defined as a network regulator and it's targets.

One key finding is that cross-tissue co-expression networks (networks composed of genes from different tissues) to be more strongly correlated with plasma protein expression compared to tissue-specific networks, including networks previously linked to cardiometabolic disease.

In this paper, we have integrated data for 974 plasma proteins with mRNA expression profiles from 7 vascular and metabolic tissues, from the STARNET study, for 532 coronary artery disease patients. In addition to discrete profiles, we also consider QTLs and, critically, transcriptional networks.

Oslo airport has a "therapy dog" at my gate for people to spend time with while waiting for their flight! Definitely a welcome distraction from the usual misery of plane travel.