Changes in Version 2.0¶

Version 2.0 brings a cleaned-up API to forgi. The BulgeGraph object should now be treated as immuteable on the level of defines and edges. Furthermore, the new Sequence class deals with many quirks of the PDB format, including tuple-indices and missing and modified residues. We now fully support the MMCIF fileformat.

Some of the biggest changes:

• All RNA file-formats can be loaded with the single function `forgi.load_rna`, which returns a list of connected components.

• The old functions for creating the RNA from a specific file type are now classmethods of the BulgeGraph class. BulgeGraph.__init__ should no longer be used.

• The API of the Bulgegraph object was cleaned up, including the following changes:

  • Functions modifying the defines and edges of the BulgeGraph (like add_node) have been moved to the private module forgi.graph._graph_construction, as they are only used during creation of the BulgeGraph object, including dissolve_length_one_stems. Instead, length_one_stem_basepairs was added.
  • To get a modified copy of a BulgeGraph, use the member methods of the GraphTransformer, accessed with .transformed. These methods are implemented in forgi/graph/transform_graphs.py and more methods will be added in the future.
  • Some functions that are probably not useful for client code have been made private or removed, including:

  • all_connections -> _all_connections
  • compare_stems, compare_*: removed
  • find_external_loops, use get_domains
  • flanking_nucleotides, use define_a.
  • get_any_sides, get_sides_plus removed
  • Functions only called by to_bg_string are now private (get_*_string, e.g. _get_connect_str).
  • get_stem_direction removed
  • get_vertex removed
  • nd_define_iterator removed
  • subseq_with_cutpoints removed, use Sequence class’s __getitem__

  • get_node_from_residue_num is deprecated and shuld be replaced by get_elem
  • Better support for forgi.graph.residue.RESID as indices.
  • New properties/ methods junctions, rods, pseudoknotted_basepairs, seq_length

• A new sequence class, which allows for indexing with integers (1-based, like the defines) and forgi.graph.residue.RESID tuples. Conversion between these two indices is now also delegated to this class. __getitem__ is implemented to properly allow for slices with cutpoints.

• Some changes were made to the CoarseGrainRNA class:

• Virtual residues are now stored for loops (at the C1’ atom), if the RNA was loaded from a PDB or MMCIF file.
• Experimental functions are now made private.
• New function get_incomplete_elements for elements with missing residues.
• New function rotate_translate

• PDB files can now be read, even if neither DSSR nor MC-Annotate are installed. However, the accurracy of the secondary structure may suffer in this case.
• forgi.threedee.model.similarity.cg_rmsd now takes virtual residues of loops into account.
• The module forgi.graph.numbered_dotbracket allows you to modify the dotbracket structure while keeping track of the mapping between brackets and residues.

Changes in Version 1.1¶

• We now support DSSR as an alternative to MC-Annotate. See the Installation for more information
• RESID for pdb-derived residue ids in now moved into a seperate module.
• describe_cg.py now adds the number of the multiloop in the sequence to the generated csv.
• Bugfixes:
  • python3 compatibility of visualize_cg.py
  • windows-compatibility of visualize_cg.py
  • fix count of multiloop-types in describe_cg.py
  • persist cofold-breakpoints in cg-files, even if no sequence was given.
  • fix bugs in parsing of MC-Annotate output. Forgi should now be able to load most pdb-files.
  • properly propagate the --keep-length-one-stems option.

Changes in Version 1.0¶

With version 1.0, we introduce support for cofold and multifold structures. Further more, we are finally compatible with Python 3 and Python 2

In the following list of the biggest changes, the most important changes are highlighted in bold.

• Python 3 compatibility: forgi is now compatible with python 2 and 3. In paticular, we have automatic tests for python 3.5, 3.6 and 2.7

• Logging support. We now use the logging module instead of print statements.

• In forgi.graph.bulge_graph:

  • Cofold Structures A BulgeGraph can now hold multiple RNA chains,

    if they are all connected by basepairs.

  • bg.seq_ids now holds named tuples (chain, resid)

  • New Errors (GraphConstructionError and GraphIntegrityError) as specializations of ValueError are introduced and sometimes raised by BulgeGraph instances.

  • Fasta sequences containing ‘T’ characters are now allowed (and converted to ‘U’)

  • Sequence class: Introduced Sequence class, a string subclass with 1-based indexing and support for ‘&’ (cofold structures)

  • define_range_iterator no longer accepts the seq_ids keyword.

  • The Bulge-graph now has a new method define_a. It returns the define with adjacent nucleotides (if present) and also works for 0-nucleotide multiloops.

  • Consistent element numbering:

    • During graph creation, multiloops are now numbered from m0, m1, … according to their position along the backbone.
    • The numbering of elements during bulge-graph construction is now consistent independent of the order of dictionaries (which was randomized in python 3 and is ordered in python 3.6)

  • Looking at multiloops as a whole:

    Multiloops continue to consist of independent multiloop segments. However, we additionally introduced methods to look at multiloops as a whole.

    • find_mlonly_multiloops finds out, which multiloop segment belong

      together in a junction

    • describe_multiloop: reports whether the junctions found with the

      previous method belong to a pseudoknot, a multiloop or an exterior loop.

  • get_angle_type now supports the allow_broken keyword to return an angle type instread of None in case of Multiloop segments that are not part of the Minimum Spanning Tree

  • bg.iter_elements_along_backbone introduced.

• forgi.threedee.utilities.average_atom_positions was converted to a JSON

  data file, which is read by forgi.threedee.utilities.graph_pdb. This saves time upon import. In the future, average_atom_positions might be entirely removed.

• New module forgi.threedee.model.linecloud to store the coordinate and twist vectors. This module’s classes implement a Mapping interface but internally use a numpy array, which allows for speedup of many operations.

• In forgi.graph.bulge_graph:

  • Exceptions CgConstructionErrorCgConstructionError and CgIntegrityError as subclasses of the new bulge graph exceptions
  • Load all chains from a PDB file: The module level function connected_cgs_from_pdb loads all RNA chains from a PDB file and greates a cg object for each connected component.
  • Bugfix: Supplying the secondary structure during loading of PDBs now works again.
  • Convenience function cg.get_stats as a wrapper around get_bulge_abgle_stats, get_loop_stats and get_stem_stats respectively.
  • Code for steric value and stacking detection is EXPERIMENTAL!

• Modules ftm.ensemble, ftm.ensemble2 and ftu.dssr are EXPERIMENTAL

• Faster RMSD by using QCOrot algorithm with Cython instead of Kabsch algorithm

• AngleStat class now supports unary minus operator.

• Recalculated average_stem_vres_positions from the NR-list.

• In forgi.threedee.utilities.graph_pdb: virtual stats and sum of angle stats

  to describe orientation of non-adjacent stems in multiloops.

• Modified Residues: Better treatement of modified residues. We try to query PDBeChem to replace the modified residue with the unmodified parent.

Changes in Version 0.4¶

• In forgi.graph.bulge_graph:

  • Speed improvement: Basepair distances between elements are cached.
  • The Bulge-graph object and file format supports arbitrary key-value pairs in the info dictionary.
  • BulgeGraph.get_connected_nucleotides no longer sorts the output nucleotides. Now this function depends on the order of stem1 and stem2 and can thus be used to determine the direction of a bulge. This is used in the new member function get_link_direction.
  • Added function BulgeGraph.get_domains, to return a lists of pseudoknots, multiloops and rods. The interface of this function might change in the future.
  • Merged pull-request by tcarlile for forgi.graph.bulge_graph:
    • BulgeGraph.get_stem_edge(self, stem, pos): Returns 0 if pos on 5’ edge of stem, returns 1 if pos on 3’ edge of stem.
    • BulgeGraph.shortest_path(self, e1, e2): Returns a list of the nodes along the shortest path between e1, and e2.

• Restructured forgi.threedee.model.comparison and forgi.threedee.utilities.rmsd into forgi.threedee.model.similarity and forgi.threedee.model.descriptors The similarity module contains all functions for the comparison of two point clouds or two cg structures. The descriptors module contains functions for describing a single point cloud, such as the radius of gyration or new functions for the gyration tensor.

• average_stem_vres_positions are back with recalculated values

• Changes in forgi.threedee.model.coarse_grain to the CoarseGrainRNA object:

  • In the self.coords dictionary, the start and end coordinate are now in a consistent order.

  • Call new member function self.add_all_virtual_residues instead of forgi.threedee.utilities.graph_pdb.add_virtual_residues

  • Coordinates of interior loops and multiloop segmentsa are no longer stored in the cg-files, as they can be deduced from stem coordinates.

    • New member function self.add_bulge_coords_from_stems is provided instead of function forgi.threedee.utilities.graph_pdb.add_bulge_information_from_pdb_chain

  • Member function self.get_virtual_residue(pos) is provided for easier access than directly via self.v3dposs. For single stranded regions, virtual residue positions along the direct line of the coarse grain element can be estimated optionally. Virtual residues are cached and the cache is automatically cleared when the coordinates or twists of the coarse grained RNA are changed.

  • Functions for creating coordinate arrays for the structure

    • self.get_ordered_stem_poss for the start and end coordinates of stems.
    • self.get_ordered_virtual_residue_poss for all virtual residue coordinates of stems. Replaces forgi.threedee.utilities.graph_pdb.bg_virtual_residues
    • self.get_poss_for_domain for coordinates only for certain coarse grained elements.

  • Removed the addition of a pseudo-vector to loop stats in get_loop_stats, which was used to avoid zero-length bulges. Now 0-length bulges are possible. This makes saving and loading stats consistent.

  • self.get_coordinates_array now returns a 2D nx3 numpy array holding n coordinate entries. You can use numpy’s .flatten() to generate a 1D array. If you want to load a 1D flat coordinate array a, use self.load_coordinates_array(a.reshape((-1,3))

  • Overrides the newly introduced method sorted_edges_for_mst from BulgeGraph. Now elements that have no sampled entry are broken preferedly. This should ensure that the minimal spanning tree is the same after saving and loading an RNA generated with the program Ernwin to/from a file.

  • self.coords_to_directions and coords_from_directions: Export the coordinates as an array of directions (end-start). This array has only 1 entry per coarse grained element.

• In forgi.threedee.model.stats: Added class for clustered angle stats.

• Changes in forgi.projection.hausdorff.

• Changes in forgi.projection.projection2d

  • Faster rotation and rasterization.
  • selected virtual residues can be included in the projection

• In forgi.threedee.utilities.graph_pdb: Added functions get_basepair_center and get_basepair_plane. This will be used in the future for stacking detection.

Changes in Version 0.3¶

• CoarseGrainRNA now has a member cg.virtual_atoms() which is used for caching of virtual atom positions. forgi.threedee.utilities.graph_pdb.virtual_atoms() now only calculates atom positions when they are needed. The two changes together lead to a great speed improvement.
• The subpackage forgi.visual was started for easy visualization of RNA using fornac or matplotlib. This subpackage is in an early development stage and will be improved in future versions.
• The subpackage forgi.projection was added to work with projections of CoarseGrainedRNA objects onto a 2D plane.
• Now forgi.threedee.model.average_atom_positions is used for all virtual atom calculations while average_stem_vres_positions has been removed. This leads to more consistent virtual atom calculations. Further more the values in average_atom_positions have been recalculated.
• In forgi.threedee.model.rmsd, the functions centered_rmsd and centered_drmsd have been deprecated and deleted respectively. Use rmsd(cg1,cg2) for a centered RMSD. This removes code duplication.
• In forgi.threedee.model.comparison a ConfusionMatrix class was introduced for speedup with repeated comparisons to the same reference.
• Several smaller changes and improvements