Comparing three methodologies for network analysis of human [11C]glyburide whole-body PET data: d-networks, s-networks, and ΔPCC networks

Abstract Background Dynamic whole-body PET and total-body PET both supply large datasets that include multiple organs, opening the opportunity to study systems biology via appropriate analysis. Network analysis, commonly used with brain imaging, is applied here with whole-body PET to compare data from different tissues and subjects before and after precipitating a pharmacokinetic drug-drug interaction. This is done with [11C]glyburide PET, a radiotracer whose tissue distribution is mediated by organic anion-transporting polypeptides (OATP) transporter function. OATPs control the uptake of drugs, primarily into the liver. We examine three methods of network analysis to evaluate the effort, efficacy, and potential applications for further research use. This was performed with 22 dynamic [11C]glyburide whole-body PET scans of healthy humans. This includes 13 baseline scans, and 9 after infusion of rifampicin, a potent OATP inhibitor. All methods use Pearson correlation coefficient. The first method generates “d-networks” using dynamic data from one region. The second method generates “s-networks” using static data from multiple regions, but not all regions. The final method generates “ΔPCC networks” with static data from all regions. The first two methods compare subjects, whereas the third compares regions. Results The d-network differentiates control and rifampicin subjects within the primary region of interest, the liver, but the differentiation is not complete, and there is still some correlation across subjects (r>0.65, p0.65, p0.18, p