If you want a DFG out like conformation for a kinase were that conformer is not available from an empirical source your basic options are:

Loop (re)Modeling or Homology Modeling from an appropriate (set of) template(s).  with Schrodinger, the underlying engine for these is Prime (You won't be able to do it with the just the free version of Maestro).  IF you do have it go through the manual, look at the knowledge base, or email them for help.  For the rest of this I'll be giving the more broadly useful general info.

Many of the family members tend away from the DGF out as it has a notably higher delta G, so you may need to scan a number of variations, in addition, while you most likely will prefer to keep as much of the structure the same as possible it will probably be necessary to remodel a notable portion of the adjacency at a time.

There are also number of web servers for this such as:

http://modbase.compbio.ucsf.edu/modloop/

If you choose to attempt homology modeling, consider mixing dedicated templates ex. if you are modeling a SRC family member that does not have an identified DFG out,  Call it UNK, and UNK has >60 % ID to SRC, and >80% similarity and an overall  Calpha RMSD < 3.0.  You could make a two template model:

Template1  maps UNK residues with 2 or more heavy atoms within N angstroms of the DGF to a SRC-DFGout crystal structure.

Template2 maps everything else to the best available structure (presumably the UNK)

free programs that do this include Modeller, tasser, and rosetta, (I believe tasser, allows this level of control on their web server as well http://zhanglab.ccmb.med.umich.edu/I-TASSER/)

Both of these approaches will have their problems and need checking - keep an eye on the Ramachandran plots (phi/psi angles), and consider following up with Energy minimization if not full MD convergence checks

good luck

1. There are quite a few publications regarding DFG-in and DFG-out homology models.

Ung and Schlessinger 2014 (http://pubs.acs.org/doi/abs/10.1021/cb500696t)

Xu et. al. 2011 (http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0022644)

Podlipnik et. al 2010 (http://www.sciencedirect.com/science/article/pii/S1093326310001439)

2. Pocketome.org integrates highresolution RTK structures.

3. If you're interested in simulating the DFG-in and out conversions, an exhaustive sampling is mandated. Either replica-exchange umbrella sampling, metadynamics, or micro-second MD using ANTON are desirable.

Thanks Jacob Pessin and Jiyong Park.