Improved, Best-like contact analysis (Issue #702) #708

jandom · 2016-02-09T22:00:47Z

Addresses #702.

A worked example with GpA dimer in a bilayer

Added tests to the BestHummerContacts calculation

Tests for fraction of native contacts (Q)

A worked example with GpA dimer in a bilayer Added tests to the BestHummerContacts calculation Tests for fraction of native contacts (Q)

hainm · 2016-02-09T22:49:24Z

package/MDAnalysis/analysis/contacts.py

@@ -146,14 +144,15 @@
 import errno
 import warnings
 import bz2
-from six.moves import zip
+from itertools import izip


should keep from six.moves import zip

zip here is izip in py2

the six import was correct. It does import the iterator version of zip under python2 as zip. So no more izip. This is part of our effort to support python3 along with python2 in the future.

- removed a dependancy on pandas, update tests

jandom · 2016-02-10T17:13:50Z

Ah, crap - it seems that I applied my changes to some ghetto-old version of contacts.py. This is now changed: the only part of contacts.py that differs from the version in develop branch.

So basically all your comments about contacts.py are valid but to some old version of contacts.py

The only thing that I really added is:

class BestHummerContacts(AnalysisBase), object-oriented
calculate_contacts, function

richardjgowers · 2016-02-10T17:21:24Z

package/MDAnalysis/analysis/contacts.py

+        step : int, optional
+            Step between frames to analyse, Default: 1
+        """
+        self.u = u


You should be able to pull the reference to universe off grA to reduce the number of arguments here

Agreed, provided that

assert(grA.universe == grB.universe)

I prefer passing groups instead of atom selection string. Whenever i pass a selection string to rms_fit_trj and i cock something up (selections are not identical, etc) i takes an extra step to debug it (gotta create the AtomGroups by hand). Plus people may be constructing AtomGroups by hand in this scenario.

jandom · 2016-02-10T17:54:23Z

@richardjgowers many thanks for all your comments and helping me navigate in the conventions! The tests fail because I use a function that depends on scipy - that used to be a dep of MDAnalysis, what's the replacement now?

======================================================================

ERROR: Failure: ImportError (No module named scipy)

----------------------------------------------------------------------

Traceback (most recent call last):

  File "/home/travis/miniconda/envs/pyenv/lib/python2.7/site-packages/nose/loader.py", line 418, in loadTestsFromName

    addr.filename, addr.module)

  File "/home/travis/miniconda/envs/pyenv/lib/python2.7/site-packages/nose/importer.py", line 47, in importFromPath

    return self.importFromDir(dir_path, fqname)

  File "/home/travis/miniconda/envs/pyenv/lib/python2.7/site-packages/nose/importer.py", line 94, in importFromDir

    mod = load_module(part_fqname, fh, filename, desc)

  File "/home/travis/miniconda/envs/pyenv/lib/python2.7/site-packages/MDAnalysisTests/analysis/test_contacts_best.py", line 23, in <module>

    from MDAnalysis.analysis.contacts import BestHummerContacts, calculate_contacts

  File "/home/travis/miniconda/envs/pyenv/lib/python2.7/site-packages/MDAnalysis/analysis/contacts.py", line 859, in <module>

    from MDAnalysis.analysis.distances import distance_array

  File "/home/travis/miniconda/envs/pyenv/lib/python2.7/site-packages/MDAnalysis/analysis/distances.py", line 34, in <module>

    from scipy import sparse

ImportError: No module named scipy```

richardjgowers · 2016-02-10T17:57:01Z

package/MDAnalysis/analysis/contacts.py

+        results: list 
+            Fraction of native contacts for each frame
+        """
+        assert(grA.universe == grB.universe)


You don't want an assert here, do a proper if not x==y: raise ValueError with a nice message

richardjgowers · 2016-02-10T18:05:09Z

I think scipy is optional now, not sure there's a workaround for sparse matrices. You can decorate the test in a skipif (grep for another scipy test in analysis), we run 2 lots of tests, minimal and full.

…e import

jandom · 2016-02-10T20:57:47Z

Build passes, allelujah!

orbeckst · 2016-02-11T06:18:29Z

backlink to #702

kain88-de · 2016-02-24T21:26:33Z

Thanks for giving me some time

jandom · 2016-02-24T21:27:54Z

@kain88-de @orbeckst all done here again (I think...) added numpy style docs, reworked the test, fixed the staticmethod problem

@jandom

- ContactAnalysis was removed by @jandom but needs to stay because the functionality does not exist in the new Contact class (But it needs to be rewritten with AnalysisBase.) - updated doc strings (numpy style)

…o issue-702-contacts

orbeckst · 2016-02-25T00:29:55Z

@jandom, you have a PR jandom#1 for the remaining stuff.
Note that I put ContactAnalysis back in – it's terribly ugly but that's something for another issue.

orbeckst · 2016-02-25T00:36:47Z

package/MDAnalysis/analysis/contacts.py


 logger = logging.getLogger("MDAnalysis.analysis.contacts")

-
-class ContactAnalysis(object):


I put ContactAnalysis back in jandom#1 – it should not have been removed. Raise an issue to overhaul ContactAnalysis to work within the new analysis model. We can then also rename it and break it's API, or for all I care, completely rewrite with more focused functionality. But for right now it has to stay.

orbeckst · 2016-02-25T00:37:52Z

@jandom , add an entry to CHANGELOG!

jandom · 2016-02-25T15:29:28Z

Aye, no problem adding back ContactAnalysis - will update the CHANGELOG

orbeckst · 2016-02-26T06:13:06Z

@jandom, did you merge my PR?

jbarnoud · 2016-02-26T12:28:05Z

package/MDAnalysis/analysis/contacts.py

+        self.outfile = outfile
+
+    @staticmethod
+    def load(filename):


This method uses self, it cannot be a static method.

…alysis into orbeckst-issue-702-contacts Conflicts: package/MDAnalysis/analysis/contacts.py

orbeckst · 2016-02-26T19:47:32Z

This looks done to me. I added the WIP flag so that it doesn't get merged for 0.14.x (as discussed with @kain88-de ).

That was a long process but along the way we made big strides towards a better analysis module #719 . Many thanks to @jandom and everyone else involved in the discussion.

- numpy style docs - guard scipy.sparse import and warn - see PR #708 for discussion

kain88-de · 2016-02-29T09:34:44Z

I merged it by hand see 4ee46b.

jandom · 2016-02-29T14:58:28Z

@kain88-de ditto #702 :)

orbeckst · 2016-02-29T15:23:10Z

If this is now on develop then the current develop will have to become 0.15.0. If we want a 0.14.1 then we need to branch from before this merge.

With semantic versioning we can't have new features in a patch release.

Oliver Beckstein
email: [email protected]

Am Feb 29, 2016 um 1:34 schrieb kain88-de [email protected]:

I merged it by hand see 4ee46b.

—
Reply to this email directly or view it on GitHub.

kain88-de · 2016-02-29T15:30:47Z

There isn't much for the patch release. We should be able to cherry-pick them into a release-0.14.1 branch. Besides I'm good with just cutting a new release every 1-2 months. That is very predictable for me and enough time to get 1 or 2 bigger things merged.

kain88-de · 2016-03-02T19:34:45Z

I've started working on improvements of the new Contacts class, you can have a look here.

I want to add some additional examples and documentation to the module. For this it would be nice to know what other people want to use the contact analysis for.

Use Cases

native contacts for folding
native contacts relative to two conformations ("q1-q2") for tracking conformational transitions (see Fig 7A in @sseyler 's PSA paper)
contacts in a binding pocket between two proteins
I read something about centroid with membrane sidechains

It would be nice if others contribute as well what this type of analysis can be used for. That way I can write examples and add tests for all of them (hopefully).

History

2016-03-02: added q1-q2 as example — @orbeckst

orbeckst · 2016-03-02T20:57:13Z

Actually, let's have this discussion in a new issue!

- numpy style docs - guard scipy.sparse import and warn - see PR MDAnalysis#708 for discussion

Improved, Best-like contact analysis (Issue MDAnalysis#702)

755631c

A worked example with GpA dimer in a bilayer Added tests to the BestHummerContacts calculation Tests for fraction of native contacts (Q)

hainm reviewed Feb 9, 2016
View reviewed changes

Revert to changes to contacts.py

cd67662

- removed a dependancy on pandas, update tests

richardjgowers reviewed Feb 10, 2016
View reviewed changes

Arguments refactored, numpy-style docs

a55119a

richardjgowers reviewed Feb 10, 2016
View reviewed changes

Jan Domanski added 2 commits February 10, 2016 14:17

Refactor universe identity check between groups, optional scipy.spars…

093a091

…e import

Add test files to package, rename test class for contacts

c6fe71e

doc string upgrades in analysis.contacts

d5e2995

orbeckst added 2 commits February 24, 2016 16:53

added back ContactAnalysis

03980b9

- ContactAnalysis was removed by @jandom but needs to stay because the functionality does not exist in the new Contact class (But it needs to be rewritten with AnalysisBase.) - updated doc strings (numpy style)

Merge branch 'issue-702-contacts' of github.com:jandom/mdanalysis int…

a2a3379

…o issue-702-contacts

orbeckst reviewed Feb 25, 2016
View reviewed changes

Upated CHANGELOG, brought back the old ContactAnalysis class

c82c185

jbarnoud reviewed Feb 26, 2016
View reviewed changes

Jan Domanski added 3 commits February 26, 2016 10:44

Remove any self calls from load method, to make it nice and static

2e7c4ee

Merge branch 'issue-702-contacts' of https://github.com/orbeckst/mdan…

d8df641

…alysis into orbeckst-issue-702-contacts Conflicts: package/MDAnalysis/analysis/contacts.py

Merge branch 'orbeckst-issue-702-contacts' into issue-702-contacts

395a41a

orbeckst added the Work in progress label Feb 26, 2016

orbeckst mentioned this pull request Feb 26, 2016

prescribe to only use top level imports in MDAnalysis.analysis #666

Closed

richardjgowers mentioned this pull request Feb 27, 2016

mandate tests for additions to MDAnalysis.analysis and introduction of analysis.legacy #743

Closed

2 tasks

kain88-de pushed a commit that referenced this pull request Feb 29, 2016

improvements for distances

0b31850

- numpy style docs - guard scipy.sparse import and warn - see PR #708 for discussion

kain88-de closed this Feb 29, 2016

orbeckst removed the Work in progress label Mar 2, 2016

backpropper pushed a commit to backpropper/mdanalysis that referenced this pull request Mar 13, 2016

improvements for distances

d5fe355

- numpy style docs - guard scipy.sparse import and warn - see PR MDAnalysis#708 for discussion

kain88-de mentioned this pull request Mar 20, 2016

Contact improvements #792

Merged

4 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improved, Best-like contact analysis (Issue #702) #708

Improved, Best-like contact analysis (Issue #702) #708

jandom commented Feb 9, 2016

hainm Feb 9, 2016

kain88-de Feb 9, 2016

jandom commented Feb 10, 2016

richardjgowers Feb 10, 2016

jandom Feb 10, 2016

jandom commented Feb 10, 2016

richardjgowers Feb 10, 2016

richardjgowers commented Feb 10, 2016

jandom commented Feb 10, 2016

orbeckst commented Feb 11, 2016

kain88-de commented Feb 24, 2016

jandom commented Feb 24, 2016

orbeckst commented Feb 25, 2016

orbeckst Feb 25, 2016

orbeckst commented Feb 25, 2016

jandom commented Feb 25, 2016

orbeckst commented Feb 26, 2016

jbarnoud Feb 26, 2016

orbeckst commented Feb 26, 2016

kain88-de commented Feb 29, 2016

jandom commented Feb 29, 2016

orbeckst commented Feb 29, 2016

kain88-de commented Feb 29, 2016

kain88-de commented Mar 2, 2016

orbeckst commented Mar 2, 2016


		logger = logging.getLogger("MDAnalysis.analysis.contacts")


		class ContactAnalysis(object):

Improved, Best-like contact analysis (Issue #702) #708

Improved, Best-like contact analysis (Issue #702) #708

Conversation

jandom commented Feb 9, 2016

hainm Feb 9, 2016

Choose a reason for hiding this comment

kain88-de Feb 9, 2016

Choose a reason for hiding this comment

jandom commented Feb 10, 2016

richardjgowers Feb 10, 2016

Choose a reason for hiding this comment

jandom Feb 10, 2016

Choose a reason for hiding this comment

jandom commented Feb 10, 2016

richardjgowers Feb 10, 2016

Choose a reason for hiding this comment

richardjgowers commented Feb 10, 2016

jandom commented Feb 10, 2016

orbeckst commented Feb 11, 2016

kain88-de commented Feb 24, 2016

jandom commented Feb 24, 2016

orbeckst commented Feb 25, 2016

orbeckst Feb 25, 2016

Choose a reason for hiding this comment

orbeckst commented Feb 25, 2016

jandom commented Feb 25, 2016

orbeckst commented Feb 26, 2016

jbarnoud Feb 26, 2016

Choose a reason for hiding this comment

orbeckst commented Feb 26, 2016

kain88-de commented Feb 29, 2016

jandom commented Feb 29, 2016

orbeckst commented Feb 29, 2016

kain88-de commented Feb 29, 2016

kain88-de commented Mar 2, 2016

Use Cases

History

orbeckst commented Mar 2, 2016