Release 0.14.0

We have just released MDAnalysis version 0.14.0. This release contains a large number of new features and bug fixes. The highlights are listed below but for more details see the release notes.


You can upgrade with pip install --upgrade MDAnalysis

Noticable Changes

Implicit OR in selections

Many long selection strings used in select_atoms have been simplified through allowing implicit OR in the arguments. For example to select all atoms with one of a few names previous required lots of ORs

u.select_atoms('name Ca or name N or name Ch')

u.select_atoms('name Ca N Ch')

The new syntax allows multiple arguments after the keyword name. The selection will keep eating arguments till it hits a keyword. The use of wildcards is still possible too, making the selection of all atoms with a type beginning with ‘C’ or ‘N’ as simple as:

u.select_atoms('type C* N*')

Similarly, for selecting ranges of resids

u.select_atoms('resid 1:10 or resid 40:50 or resid 56 or resid 67')

u.select_atoms('resid 1:10 40:50 56 67')

This new behaviour works for name, type, resname, segid, altLoc, resid, resnum and bynum selections! The old behaviour will still work, but we feel this should save a lot of typing!

Boolean indexing of trajectories

You can now treat trajectories like numpy arrays in a fully pythonic fashion: You could already do fancy indexing and now you can also do boolean indexing. One application is building simple re-usable trajectory filters with ease, e.g., in order to slice data.

# build a filter array for re-use: only frames where the centroids
# of groups A and B are closer than 3 A
idx = [np.linalg.norm(A.centroid() - B.centroid()) < 3.0 for ts in u.trajectory]

for ts in u.trajectory[idx]:
   # analyze the frames where |A - B| distance < 3

for ts in u.trajectory[np.logical_not(idx)]:
   # do analysis on the other frames, |A - B| >= 3

New analysis module to calculate linear densities

The MDAnalysis.analysis.lineardensity module contains the class LinearDensity that simplifies calculation of mass and charge densities profiles along the primary axes of a simulation cell.

Rewrite of TRR and XTC file handling

Our wrapper of the Gromacs library xdrlib has been completely rewritten in cython. This changes brings us one step closer towards supporting Python 3. The only user facing API change is that we don’t save persistent frame offsets with the pickle module anymore but with numpy’s ‘npz’ format. This improves reopening of xtc/trr files.

Experimental Python 3 Support

With this release it is possible to run MDAnalysis under Python 3. We have ported most of our coordinate readers, DCD is still missing, and topology readers to be python2/3 compatible. Most features should already work but expect there to be some minor glitches.


This release contains other performance enhancements and fixes. For a detailed list see the release notes.

Release 0.13.0

We have just released MDAnalysis version 0.13.0.


You can upgrade with pip install --upgrade MDAnalysis

Noticable Changes

Usage of openMP

The MDAnalysis.lib.parallel.distances module has been merged into MDAnalysis.lib.distances. To select the openMP implementation all functions in that module now accept a backend='openMP' keyword. See #530.

Performance Enhancements for GRO-files

Reading GRO files has been sped up for large files. Velocities will now also be written written for GRO files if possible.


This release contains other performance enhancements and fixes. For a detailed list see the release notes

The benefits of social coding

MDAnalysis has truly arrived in the 21st century: we now even have a blog (in addition to a Twitter account @mdanalysis) and of course, development occurs on GitHub, the “social coding platform” (at least that was the tag line a while back). However, for most of its life (January 2008 – March 2015) MDAnalysis lived on Google Code and the move to GitHub was done more out of necessity (because of Google Code’s shutdown) than for any other reasons. After all, we knew, moving a repository, issue tracker, wiki, and what-not would be painful and what could we possibly gain?

In turns out, a lot.

MDAnalysis commit history (Jan 2008 to Sept 2015)

As you can see from the image of the commit history, moving to GitHub in April 2015 coincided with a flurry of activity and significant contributions from new developers, two of which are now also on the coredevs team. The graph, of course, does not demonstrate causation, only correlation. But anecdotally it seems to be the case that the ease with which one can communicate on GitHub through the issue tracker and comment on code through pull requests helps people to enjoy open source software development. In addition, the good integration with continuous integration services such as Travis CI, coveralls or quantifiedcode takes care of many essential but boring tasks and allows people to focus on what they like doing — writing code that helps them and others to get their research done.

Given that an open source projects such as MDAnalysis very much depends on developers offering up their own time and expertise, working on a platform that makes it (mostly) enjoyable for them, is rather valuable. And maybe, before trying it, we just slightly underestimated the social component of writing code.