Validating the Network-Based Objective Measure for Helical Membrane Protein Structures

Computational structure prediction, including de novo and homology modeling, is an important tool for membrane protein studies. Developing an accurate scoring function that can be used for structure discrimination and assessment remains a challenge. In our previous work, we have analyzed a set of high-resolution membrane protein structures using the network approach developed in our lab and proposed such a scoring function. Here, we aimed to validate this function using three diverse datasets. First, we applied it to a diverse set of 139 homology models. The models included in this dataset represent ten unique membrane protein superfamilies and have 20-100% sequence identity to their respective templates. Next, this scoring function was adopted for the analysis of nine bacteriorhodopsin structures of different resolution. Finally, the discrimination capability of the proposed scoring function was tested using the HOMEP benchmark dataset of 92 membrane protein models. Detailed analyses of the results confirmed that the proposed network measure can be adopted as an objective packing quality indicator and for structure discrimination. These results suggested that the proposed scoring function is a good indicator of high-quality models and can be applied to a variety of membrane protein models.