i've noticed something strange: let's try build serie of models in model_default.py-way, but with starting_model=1 and ending_model=10. we get 10 different models a1..a10. let's do it one more time with same script and we get another 10 models b1..b10. they are different between themself but coincide with previous serie, such as a1=b1..a10=b10. i just don't understand it. each model is calculated by automodel.single_model(), inside it method self.randomize_initial_structure(atmsel) is called. so randomization takes place. how it can be that these randomizations led to coinciding results?
and the main question - how to make modeller give really random serie of models?
kind regards, Igor.
On 10/24/11 7:30 AM, Igor Dvortsov wrote: > i've noticed something strange: let's try build serie of models in > model_default.py-way, but with starting_model=1 and ending_model=10. we > get 10 different models a1..a10. let's do it one more time with same > script and we get another 10 models b1..b10. they are different between > themself but coincide with previous serie, such as a1=b1..a10=b10. > i just don't understand it. each model is calculated by > automodel.single_model(), inside it method > self.randomize_initial_structure(atmsel) is called. so randomization > takes place. how it can be that these randomizations led to coinciding > results?
That's not strange at all - Modeller is deterministic, so will give you the same results given the same input conditions and random seed. If you want different results, change the random seed: http://salilab.org/modeller/9.10/manual/node111.html
Ben Webb, Modeller Caretaker