Changes since release 9.12

• Modeller now includes a variety of SOAP (statistically optimized atomic potential) scores for assessing proteins, loops, and interfaces.

• Bundled version of HDF5 updated to 1.8.10.

• The Lennard-Jones interaction energy is now artificially truncated at very short distance; this makes simulations with poor starting conditions much less likely to 'blow up'.

• Residue, Atom and Chain objects can now be meaningfully compared in Python (previously == would always return False).

• model.get_insertions(), model.get_deletions() and model.loops() now have an include_termini option; if False, residue ranges that include chain termini are excluded from the output.

• Most Modeller objects can now be written to or read from file handles (e.g. modfile.File(), or any Python filelike object) as well as named files.

• automodel now accepts file handles or alignment objects for the alnfile parameter.

• Structure.read() allows reading an alternative conformation of a template.

• User-defined energy terms can now report violations and energy profiles.

• The sigma factor for EM fitting can now be modified (density.sigma_factor).

• Bugfix: have Restraints.make_distance() restrain distances where the template is a non-standard residue with a standard backbone (e.g. UNK, ASX, GLX).

• Bugfix: don't falsely warn about topology being available for every BLK residue when automodel reads the initial model PDB file back in.

• Bugfix: model.get_insertions() previously ommitted the first residue of insertions in some cases, while model.get_deletions() incorrectly handled overhangs and a single residue between two deletions.

• Bugfix: model.build() no longer crashes when trying to interpolate and exactly one atom in the segment has defined coordinates.

• Bugfix: model.build() no longer uses atoms in a different chain as anchors for interpolation.

• Bugfix: derivatives of the angle feature are now more stable when the angle is close to 180 degrees.