pNNMAKE compilation

This system is deprecated (or at least, nobody who knows anything about it is willing to offer support for it). The logic of fragment selection was ported to Rosetta proper a while ago (I’m not sure if it was as early as 3.2): http://www.rosettacommons.org/manuals/archive/rosetta3.4_user_guide/dc/d10/app_fragment_picker.html

There are also some demos for it in the demos directories: rosetta34/rosetta_demos/fragment_picker/BestFragmentsProtocol

The general recommendation, if you’re not doing a massive number of runs and aren’t doing anything too custom, is to avoid all the hassle and just use the Robetta server to do fragment picking. http://robetta.bakerlab.org/

If you absolutely need to use NNMAKE to do fragment picking, I can probably track down someone to help you troubleshoot, but be advised that people have now moved over to the new fragment picker: http://www.rosettacommons.org/manuals/archive/rosetta3.4_user_guide/dc/d10/app_fragment_picker.html

I second Rocco…I’m a Rosetta developer and I almost always use Robetta instead of generating fragments myself. If it’s stock fragments, use Robetta!

Understood. I’d still recommend looking into the new fragment picker for custom jobs. As Steven indicates, the people who knew NNMAKE have either stopped using it or have moved out of the Rosetta community entirely. If you need to replicate an old protocol exactly or the like, we can possibly get NNMAKE to work, but things will likely be much easier now and in the future if you use the new fragment picker instead.

It will work the same way – you can manually edit the secondary structure prediction inputs to force the result you need. I’ve done the same thing, forcing helix in a region that SS prediction felt was loop but I had reason to believe was the C-terminus of a longer helix.

I just talked with one of the local experts, and he guesses that running the fragment picker normally will probably work. The default vall does have membrane proteins in it, and if your protein is a typical membrane protein, the protocol will likely pick those fragments out.

The one caveat is that because the vall is dominated by soluble proteins, there’s a chance that you’ll pick up spurious soluble protein fragment hits. One way around that is to trim the vall to be membrane protein specific, but that’s probably only worth doing if the default fragment picking doesn’t produce acceptable results.

The real answer is, “it’s empirical and system-dependent, you’ll have to try both to find out”.

The useful answer is, try the easiest one first and if it works, don’t bother with the other one. I suspect the ss prediction file is easier to modify than the vall.