Computational modeling including de novo protein structure prediction and homology modeling techniques have been extensively used in studies of helical membrane proteins as well as structure-based drug design efforts. Developing an accurate scoring function for structure discrimination and validation remains a challenge. Network approaches based on the overall network pattern of residue packing have been proven to be useful in soluble protein structure discrimination. In this work, we first carried out such analysis on a set of diverse and non-redundant native helical membrane protein structures using the network tool previously developed for the analysis of soluble proteins. To explore the potential application of the findings, we applied the same approach to two test sets of total 101 computationally constructed G-protein coupled receptor (GPCR) models, constructed using either de novo or homology modeling techniques. Models in these test sets have sequence identity with the bovin rhodospin varying from approximately 20% to 95%. Results of this large-scale analysis indicate that such an approach is very effective for discriminating less native membrane protein folds from native ones and the findings by studying native membrane proteins are good indicators of a native fold. These findings should be of help for the investigation of the fundamental problem of membrane protein structure prediction. This work is supported by the Research Starter Grant in Informatics from the PhRMA Foundation.