Changes since release 9.11

• Modeller is now only supported on Mac OS X 10.6 or later. If you want to run Modeller on older OS X releases, use Modeller 9.11.

• alignment.align(), alignment.salign() and alignment.align2d() now attempt to keep chain breaks aligned when aligning multi-chain sequences. This behavior can be tuned with the break_break_bonus parameter.

• alignment.edit() has a new by_chain parameter; if set, overhangs are removed for each chain in multi-chain alignments.

• alignment.edit() now returns the number of residues that it removed.

• Align codes (identifiers of sequences in alignment objects) can now be of any length. (Codes in profile and sequence_db objects are still limited to 40 characters.)

• model.build() now does a better job of generating coordinates starting from just CA positions.

• generate.transfer_xyz() now generates better structures for 6-membered rings. (Previously, initial models could occasionally contain such badly distorted rings that the optimizer was not able to recover the correct geometry in the final models.)

• Atoms in BLK residues are now assumed to have typical heavy-atom radii when calculating Lennard-Jones interactions, similar to soft-sphere (previously no interaction was calculated).

• Atom types can now be changed by assigning to Atom.type. This can be useful when employing BLK residues to use more accurate radii.

• A new function automodel.guess_atom_types() can now be called to try to assign more appropriate atom types for BLK residues.

• Modeller now matches residues in PDB files using both name and ATOM/HETATM indicator; previously just the name was used. This allows for correctly reading PDB files containing amino acid ligands, for example (they were previously forcibly mapped to ATOM on output).

• Added support for Python 3.3.

• Bundled version of HDF5 updated to 1.8.9.

• Bugfix: Q* and UNK atoms are no longer silently skipped when reading PDB files.

• Bugfix: don't weaken distance basis restraints too much when two means are too far apart, since this can cause a numeric overflow in Restraints.spline() later.