Petr Danecek wrote: > I am attempting to build a model from multiple templates, but I am > getting very bad results. After doing some tests I have discovered that > the problem lies most probably in the unwanted 3D orientation of the > blocks which the model is built from.

This is usually a symptom of incomplete optimization. You can improve this in many cases by providing a better initial model (.ini file). Also check to make sure the alignment templates have at least similar structures - otherwise you're unlikely to get a sensible consensus structure as a result. alignment.check() will usually warn about this.

> To test this, I took a single PDB file and splitted it into three chunks > aligned as: > > aaaaaaaaa------------------------- > ------aaaaaaaaaaa----------------- > --------------aaaaaaaaaaaaaaaaaaaa > aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa > > In this test, I would like to reconstruct the original structure. Is it > possible?

Very unlikely, since Modeller will build many of the distance restraints that it would do if you used the complete PDB file as a single template. For example, it cannot constrain the distance between the C and N termini any more, because they are in different templates.

> I tried to set the variable a.initial_malign3d=1, but the > modeller ends with the error "_modeller.ModellerError: fit2xyz_296E> > Number of equivalent positions < 3: 0"

This does a structural alignment between the three templates. This is very unlikely to succeed, because you have such a small overlap between them (hence the error about the number of equivalent positions being insufficient).

Ben Webb, Modeller Caretaker