update rotamer library

[orbeckst]

@katrinr thanks for updating the library. Can you please also do the following things:

1. Can you please confirm with Gunnar that we are allowed to publish the new library under the GPL (like the previous version)? I'd like to have at least an email from Gunnar stating this. If the licensing is not clear, we cannot include it here.
2. name the actual library files distinctly, i.e., have 2017 in the name of the new ones.
3. update CHANGELOG and state that we have an updated rotamer library and that this is the default now; what are the changes in the new version? If you add a description of changes I can add a new file to the online docs with a table of libraries that are included
4. update the online docs (in doc/sphinx/source – check both index.rst and background.rst)
5. update the test cases!!!

[orbeckst]

By the way, GitHub does not recognize the email [email protected] that you used for committing. Just add it to your GitHub profile as another recognized email and the display (and contribution graph will be fine).

[orbeckst]

When I run

 ./test_mtssl.sh

the script uses the new library and produces different numbers and thus the test fails.

The ./test_mtssl_pre.sh always fails for me.

We should have tests with the old and the new library.

@katrinr can you take care of the tests? Instead of directly committing, make a pull request so that we can discuss the code briefly.

Once we have tests we can also run them automatically (issue #23).

[katrinr]

Alright! :) I will take care of the testes (which only fails due to the comparison data is from the old library). I am excited to see the difference...

[katrinr]

1. Gunnar Jeschke is contaced.
2. Naming of the libraries (2011 and 2017) is implemented.
3. CHANGELOG is updated. I don't have a precise explanation of the details of the update. The new version of the rotamer library is also the default version in MMM and shows improvement in the comparison with experimental data.
4. online docs are updated
5. test cases are updated (DEER and PRE with both libraries, 2011 and 2017)

Everything is now in a new branch: update_library

[katrinr]

@orbeckst, I am working with the new library right now, but it reports that quite often no spin distances can be found. I checked the structure and aligning the rotamer library by hand, that there should be distances. Does anyone have an idea why this could be?

I decreased the clash distance and for some label pairs I obtain distances. However, also for the default clash distance I expect to see rotameric states.

Further, I changed the first rotameric state (of library 2015) to something similar as the first rotameric state of library 2011). That did not solve the problem, but gave me a different distance distribution although I changed the weights accordingly (R1A_298K_populations_2015.dat).

[lukas-stelzl]

Hi Katrin, Could you upload a test.pdb so that I could look at the problem as well? Does it also occur with the PepTSo pdb that comes with the package? Thanks

…

On Mon, Oct 9, 2017 at 4:01 PM, katrinr ***@***.***> wrote: @orbeckst <https://github.com/orbeckst>, I am working with the new library right now, but it reports that quite often no spin distances can be found. I checked the structure and aligning the rotamer library by hand, that there should be distances. Does anyone have an idea why this could be? I decreased the clash distance and for some label pairs I obtain distances. However, also for the default clash distance I expect to see rotameric states. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#21 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ATTjTtEsYdmwyfzq49KjvuCEN48_25TVks5sqicXgaJpZM4PwraR> .

[katrinr]

Thanks, @lukas-stelzl!
When I visualize the neighboring residues as spheres, I seems a bit crowded (e.g. for labeled residue 292). However, I would expect rotamaric states.

@orbeckst, as far as I can see, the code does not consider H coordinates?

cc0_2.pdb.zip

[katrinr]

Figured out the problem: the oxygen atom of MTSL is always to close to CA of residue+1 in the polypeptide chain.
I also extended the proton selection, as I had the impression it changed the hydrogen selection. The function, which is changed is also in the updated_library branch.
Thanks :)

def find_clashing_rotamers(self, fitted_rotamers, protein, site_resid):
        """Detect any rotamer that clashes with the protein."""
        # make a KD tree of the protein neighbouring atoms
        proteinNotSite = protein.select_atoms("protein and not (name H* or name [123]H or type H) "
                                              "and not (resid " + str(site_resid) + 
                                              " or (resid " + str(site_resid-1) + " and (name C or name O)) "
                                              "or (resid " + str(site_resid+1) + " and (name N or name CA)))")
        proteinNotSiteLookup = KDNS.AtomNeighborSearch(proteinNotSite)

        rotamerSel = fitted_rotamers.select_atoms("not (name H* or name [123]H or type H)")

        rotamer_clash = []
        for rotamer in fitted_rotamers.trajectory:
            bumps = proteinNotSiteLookup.search(rotamerSel, self.clashDistance)
            rotamer_clash.append(bool(bumps))
        return rotamer_clash, np.sum(rotamer_clash)

[katrinr]

The test runs don't need an update, because for peptso it gives the same result. The new selections are useful for protein dependent coordinates and pdb dependent formats (as it was the case for my test case).

[orbeckst]

Closed with #24