Another removed entry, this time  a mismatch of titles PDB vs CIF

Removed another strange entry (after july 2011 remediation) from test list

Removed a strange entry (after july 2011 remediation) from test list

Changed behaviour of getId() in ChainInterfaceList, now uses the same ids as the interfaces, i.e. starts counting from 1 and not from 0.
New graph classes for storing interface topology

Fixed bug: was calculating ASAs/BSAs with Hydrogen atoms when they were present. Calculating without them is more common practice and most importantly makes more comparable prots with/without hydrogens. Now removing Hydrogens before interfaces calculation

Improvements mainly to ChainInterface and associated classes: now can detect a lot more specific interactions in interface: disulfides, hydrogen bonds, any other kind of close interactions

Some improvements and fixes to enumerateInterfaces

Now can calculate interfaces also for non-polymer chains, introduced new parameter to do only polymer chains or both poly/non-poly.

Introduced some more checks in PDB file parser, now should be even more robust with badly formatted files.
Added some more atom types.

Fixed some issues with writing of pdb files and wrote a test for write/read pdb files.

Now using hetatoms for testing interfaces against pisa

MAJOR CHANGE: now able to read and treat properly HETATMS and nucleotides from PDB entries. All tests pass. Anyway surely there will be some bugs to iron out still.

Another test

MAJOR INTERFACE CHANGE: the main PDB data loading interface has changed. Now all loading of data occurs through PdbAsymUnit. Refactoring of some classes: Pdb is now PdbChain.

More optimisation on geometric hashing algorithm.

Now parsing pdb titles in PDB, CIF and pdbase. Tests pass.
Moved some constants from Pdb to PdbAsymUnit

Changing file names of contact map test files, windows didn't like the ':' in the file names

Another fix to PdbfilePdb parser. Now it is more robust with bad formatting, tested on a whole tar ball of CASP predictions. Added the CASP predictions test set to tests.

Now the re-numbering of residue serials in PDB files has been improved. Instead of aligning the sequences whenever the serials don't match SEQRES, we first try to see if there is a shift in the serials with respect to SEQRES (happens in many cases). Only if we can't find a shift that fully matches SEQRES we proceed to re-aligning. We also treat properly now the ins codes. Mapping is as close to CIF as possible. The parsers test passes for the whole cullpdb20. For some entries like 2ofz the mapping does not coincide 100% with CIF as there are ambiguities in the alignment.

Important change: now the PDB file reader will try to read and fix the numbering of PDB files. Whenever the alignment is wrong it will realign and renumber using the jaligner package. The result of reading original PDB files now will be the same as that of reading CIF files, including proper mapping of classical PDB numbers to SEQRES residue serials (as in CIF). 
There is one change of behaviour in comparison to before: when no SEQRES present the sequence is taken to be that of ATOM lines instead of padding it with Xs.
Still in this version the re-alignment is not perfect as there are some times when ambiguities occur and they are not solved (e.g. in 2nwr where alignment in an unobserved loop can be at two possible places for a GLY). That is anyway a rather minor problem (coordinates are still fine, just the chain is not ordered correctly at 1 or 2 points) and rare (~1% of files)

Removed two exception classes that were only used privately, didn't make much sense. Now using PdbLoadException

Renamed xxxxError classes to xxxxException as they should be.

A few bug fixes and improvements in pdb data parsing.
- bug fix: in some cases exptl method field has more than 1 value (e.g. 2krl). In cif files this was causing a null pointer. We now parse it properly (taking first one as the exptl method) in both cif and pdb files
- improved very slightly the cif file parser moving out of loops the index getters
- drop fullLength as a field in Pdb, now we have only getFullLength()
- pdb file atom parsing is now column based and not regex based (hopefully will make it slightly faster)
- now parsing the element column of pdb files: now the atom type detection comes from the appropriate field in pdb/cif/pdbase. If in pdb file and not present we still try to guess it as before

Got rid (finally!) of AAinfo. Its functionality is now in ContactType and AminoAcid.

Finally got rid of all duplicated amino acid methods in AAinfo. Now everything unified in AminoAcid enum.

Now taking paths from owl_test_paths.dat file

Fixed important bug that broke the reconstruction programs. Tinker output files don't contain occupancy/bfactors. Thus we have to make their parsing optional. Added a test case for this.

Renaming all xxxxError named exceptions to xxxxException. Long overdue thing (was historical because it came from python). Learnt just recently that java does have a concept of Error too.

Fixed important bug: we didn't model the blast data properly. A hit is composed of several hsps. Now our classes following that data model properly (before we treated a hit as an hsp).

Bringing back the PISA classes for interface, molecule and residues. We store things first into them and only then convert to our own ChainInterface if needed. This is a much better approach less prone to errors. Tests pass.

Now parsing experimental method, resolution, R free and R sym.

Forgot to fix 2 problems in last commit.

Fixed bug: was still not converting well algebraic transformations to matrix: was failing for transformations with terms like "X-Y+1/3"

Fixed bug in SpaceGroupTest. Now it works and test passes.
Introduced checking for clashes in interfaces. Added it to the PdbAsymUnitTest

Fixed some bugs with parsing symop.lib and conversion of transformations. Properly testing SpaceGroups and parsing now. Still one test (transpose equal inverse) does not work for trigonal and hexagonal groups.

Fixed bug: now we properly parse from pisa both the orthonormal and the crystal coords operators.
Added a test for SpaceGroup to test for proper parsing of algebraic strings

Made conf file path a resource path so that it is readable from jar file too.

Implemented some parallelisation in ASA calculation. It doesn't scale very well (could measure x2 speed-up with 4 CPUs in a Core2 Quad) but at least we have it, now this is starting to be better than NACCESS!

Some more complete tests.

Some optimisation of the asa code, run time improved substantially.

Implemented our own ASA calculation! Taken from Bosco's python implementation, thanks again Bosco!

Some minor improvements: now populating ids of ChainInterfaces, now interface area is half to match pisa's. 
Fixed bug in getAllInterfaces tester, wasn't comparing properly in small values. 

Finally ASA/BSA comparison to PISA test passes. Have to exclude one case where slight difference interface area values led to different sorting and so comparsion doesn't work. This will still be a problem whenever that happens (whenever different sorting of PISA vs ours)

Fixed bug: NACCESS (weirdly) calculates slightly different ASA values for same molecule in different orientation (i.e. it depends on the choice of axes)!! That produces then some strange BSA values (including some negative) and slight discrepancies. Thus now we run NACCESS always for each of the 2 isolated molecules of the interface and then the complex (and so losing efficiency).

Fixed bug: the bsa values of residues of the different interfaces were being updated on the same references, we have to copy before creating ChainInterface objects in getAllInterfaces.
Implemented interface pdb file outputting in enumerateInterfaces
Still interfaces tests don't pass because of 2 problems: a) order of first/second molecule not necessarily same as PISA (and in case of same chain codes can't catch it with chain codes), b) in many cases there are many minor discrepancies, usually associated to a lot of small negative bsa values.

Fixed minor bug in test, was not comparing properly the interface count of pisa vs ours.

Fixed bug: bsa values were wrongly calculated. 
Now sort of interfaces is always descending.
Some more testing for individual residues asa and bsa values. 
Still an issue: individual bsas are not matching pisa's even though total interface areas are fine.

Now PISA interfaces using same classes as our own calculated interfaces. Removed all Pisa specific classes (except for connector and parser). Introduce new class ChainInterfaceList which also simplifies some things.
PdbAsymUnitTest passes.

Now elimination of redundant interfaces not based only on count of edges, but also on identity of each of the contacts (atom's serials, atom codes, residue serials, residue types). Added more test cases. Seems to be working fine.

Fixed bug: was null pointing in parsing NMR cif files because they don't have crystal parameter fields. Now skipping that properly. Added NMR test entry to test cases.

First fully working version (this time hopefully for real...) of interface enumeration. Works for many examples, still there are problems in many others due to area discrepancies with pisa, but those are minor things. The elimination of duplicates is based on chain codes and number of contacting atoms, it is of course possible that 2 different interfaces happen to have the same chain codes and same number of contacting atoms (but unlikely). Would need a more fine grained comparison.
New test class to compare automatically to PISA output.

Scoring:KendallsTau-method added to  Statistics

classes moved

Now parsing the unit cell and crystal space group info when reading PDB sources. New simple script for unit cell volume.

Now parsing the occupancy and bfactors from pdbfile, ciffile or pdbase. Tests passed (improved tests so that they can run straight from an pdb ftp mirror with gz files).

New SiftsConnection and SiftsFeature classes for SIFTS pdb to uniprot mapping. Implemented the HasFeatures interface in Pdb. In the future we can hopefully use that mechanism for all features (sec. strucuture, scop and so on)

Extracted constants to owl_test_paths.dat and added CASP7 fasta files to resources dir. Now BlastParsersTest should be able to run anywhere.

Fixed hard-coded resources paths broken since major restructuring of project.

refactoring: moved many many classes to more appropriate packages; created new packages owl.core.sequence.alignment, owl.core.structure.alignment, owl.core.structure.features

refactoring: adapting test packages to match new package structure