Could you provide me with the ideas bases on which the modeler build the final models having several templates as well as its multiple-structural alignment as the inputs? In my case I'd like to build model of the GPCR having 5 templates of other receptors (with known X-ray structures) having low sequence identity against my target. What will be better i) to include all structures as the templates or ii) choose only one template with the highest sequence identify to my target? In the first case we will obtain trivial result: model will looks like the template but in the second case ( model built based on the multiple templates) - what type of averaging between templates will be used by modeler to produce final model? Please provide me with some ideas and reading about this topic.