Homology models of class A GPCRs is becoming an increasingly useful tool in drug discovery. However, the rhodopsin template structure is often inadequate to accurately describe the entire range of GPCRs that can bind ligands of largely differing sizes. We have previously described a molecular dynamics (MD) based methodology to expand and refine model GPCR active sites to specifically address the problem of not being able to fit large ligands into models of certain GPCR subfamilies [Kimura, Tebben, and Langley, submitted to Proteins]. Here we present further improvements to this method and a few case studies of successful applications.