Skip to content
Oliver Beckstein edited this page Sep 26, 2015 · 3 revisions

This is a new major release of MDAnalysis. In line with our release policy, it introduces new features at all levels of the code and removes old legacy code in favour for a more consistent (and hopefully more intuitive) API. These changes may break old scripts using MDAnalysis. Major effort went into making the AtomGroup work in all situations and allow users to access segments/residues/atoms from any Atom, AtomGroup, Residue, or Segment.

Trajectory writing has been streamlined (e.g. writing of selections can be done by just supplying the appropriate AtomGroup to the Writer.write() method).

A number of bugs have been fixed and the coverage of UnitTests expanded. New scripts and improvements to the MDAnalysis.analysis module should make it easier for users to adapt MDAnalysis to their own use cases.

If you have old code then it is worthwhile looking at the Changes below in more detail as it lists changes that are known to be incompatible with pre-0.7 code.

We hope that the improvements in this release make it worthwhile to update your old code and encourage you to report any problems via the mailing list and/or the Issue Tracker.

Changes

major release 0.7.0: (includes changes that can BREAK BACKWARDS COMPATIBILITY)

Removed all deprecated code

In particular, the following methods are gone or are not found in the previous place anymore:

  • AtomGroup.principleAxes (Issue 33)
  • DCD.DCDReader.dcd_header() and DCD.DCDWriter.dcd_header() (use _dcd_header())
  • Universe.dcd (and Universe.xtc, Universe.trr...) --- from now on only Universe.trajectory is supported. WILL BREAK LEGACY CODE!
  • removed the following packages from top-level MDAnalysis name space:
    • AtomGroup, Selection: import them from MDAnalysis.core if really needed (e.g. import MDAnalysis.core.AtomGroup)
    • distances, rms_fitting: import MDAnalysis.analysis.distances or import MDAnalysis.analysis.align.rms_fitting (the actual modules still live in MDAnalysis.core but they might get moved in the future and bundled with transformations)
    • from MDAnalysis import * will only get ['Timeseries', 'Universe', 'asUniverse', 'Writer', 'collection']
  • removed copy flag from distance_array and self_distance_array: setting it to False would always give wrong results so there was no good reason keeping it around

Trajectory I/O and file formats

  • improved trajectory writing
    • MDAnalysis.Writer() factory function that provides an appropriate writer for the desired file format
    • Writer.write() accepts a Universe or a arbitrary AtomGroup (e.g. from a selection); this is much more flexible than Writer.write_next_timestep(); the DCD/XTC/TRR writer also accepts a Timestep for its write() method.
  • CHARMM CRD coordinate files
    • CRDReader added (fixes Issue 40) ... it will work for both standard and extended formats: NO special flags needed.
    • CRDWriter will now write extended crd files: NO special flags needed.
  • PDB format: By default, PDB files are read with the PrimitivePDBReader and not the Bio.PDB reader anymore because the latter can drop atoms when they have the same name in a residue (which happens for systems generated from MD simulations) The PrimitivePDBReader copes just fine with those cases (but does not handle esoteric PDB features such as alternative atoms and insertion codes that are not needed for handling MD simulation data).
    • The default behaviour of MDAnalysis can be set through the flag MDAnalysis.core.flag['permissive_pdb_reader']. The default is True.
    • One can always manually select the PDB reader by providing the permissive keyword to Universe; e.g. Universe(PDB, permissive=False) will read the input file with the Bio.PDB reader. This might be useful when processing true Protein Databank PDB files.

Changes to AtomGroups

  • AtomGroup
    • Indexing is made consistent with the way lists behave:
      1. indexing with integers returns a single Atom
      2. slicing always returns a new AtomGroup
      3. advanced slicing with a list or array returns a new AtomGroup (NEW, fixes Issue 36)
    • AtomGroup coordinates can be manipulated (translate(), rotate() and rotateby() methods; when appropriate, these methods can take AtomGroups or arrays to determine coordinates)
    • new attributes residues and segments for AtomGroup to give access to the list of residue/segment objects of the group
    • new exception NoDataError; raised when creation of an empty AtomGroup is attempted (see also Issue 12)
    • consistent representation of the Segment > Residue > Atom hierarchy : all classes related to AtomGroup now expose the attributes atoms, residues, segments that provide access to groups of the corresponding objects
  • improvements to Residue, ResidueGroup and Segment classes; documented the instant selector pseudo-attributes:
    • documented accessing residues from Segment as Segment.r<resid>; resid is 1-based -- BREAKS OLD CODE that used the previous undocumented feature (which was 0-based)
    • added SegmentGroup class
    • can write from Residue, ResidueGroup and Segment (Issue 46)
    • residue name attribute of a Segment now consistently returns a ResidueGroup (Issue 47) -- MIGHT BREAK OLD CODE
    • added documentation and examples in the doc strings
    • new special dihedral angle selections defined for Residue class to simplify analysis of backbone torsions (experimental)

Other changes

  • whitespace is no longer required around parentheses for selectAtoms() strings but the old syntax with white space still works (Issue 43)
  • new contact_matrix method for calculating contacts (Issue 30); for large (N > ~10000) coordinate arrays automatically switches to a method using a sparse matrix (slower)
  • more example scripts (e.g. for membrane analysis, trajectory writing, coordinate transformations)
  • fixed Issue 51 (distance_array() did not properly check its input and wrong results could be returned if the input was a float64 and a float32 array)

Experimental features

  • manipulation of coordinates through the translate and rotateby methods of an AtomGroup
  • a Residue can return dihedral angle selections for the protein phi, psi, and omega backbone angles

Authors

orbeckst, denniej0, tyler.je.reddy, danny.parton, joseph.goose

Clone this wiki locally