Baptiste Legrand wrote: > Thanks for your answer, I didn't know that the default deviation is 4.0. > Is it interessant to set it to 8.0 or more or the models will be too far > from the template? Is there a theorical (estimation) limit to explore > the conformationnal space without deviate totally from the template?

The deviation parameter controls the spread of initial models, not the final optimized models - if the optimization is functioning correctly, the optimized models should look like the template anyway. (However, optimization is not a 'solved' problem, which is why the initial model is the sequence threaded onto the template structure rather than, say, the extended chain or randomly placed points.) So in practice changing deviation will not have a large effect on your model generation (unless you set it to 0, in which case no randomization is done).

Ben Webb, Modeller Caretaker