Changelog¶

Release 3.4.0¶

Sidechains and BackboneLists now use bond length and angle parameters that are extracted from a non-redundant set of high resolution X-ray structures. IC data from CHARMM36 was used before. The new parameters slightly improve sidechain modelling accuracy.
Several minor bug fixes and improvements

Bugfix release: Updates OpenStructure dependency to 2.5.0 and fixes issues with Singularity container.

Check for and report non-planar rings in the modelling result. This check is also incorporated in energy minimization, potentially increasing the number of minimization steps.
Algorithms using large model databases as template libraries for modelling. Sequence search: PentaMatch, a k-mer based sequence search with extreme speed but low sensitivity (similar implementation as https://www.rbvi.ucsf.edu/chimerax/data/kmer-aug2022/kmer_search.html). Structure storage: FSStructureServer, an OMF (OpenStructure Minimal Format) based structure storage that compresses AFDB v4 (214E6 entries) to 1.4TB.
Several minor bug fixes, improvements, and speed-ups.

Remove outdated configuration scripts (Fedora 19 / ancient Mac)
Construct CB atoms if only N,CA and C atoms are present in ReconstructSidechains input structure.
Detect Python using functionality provided by CMake instead of our own code. You might have to adapt CMake flags when building ProMod3 to variables specified here: https://cmake.org/cmake/help/latest/module/FindPython.html
Several minor bug fixes, improvements, and speed-ups.

Bugfix release: reintroduces backwards compatibility in BuildRawModel. See commit 0288eced23a9a541a88005a7ed30b8f019e06226 for details

This is expected to be the last release supporting Python 2.
This project now requires a C++11 compatible compiler.
Introduced VINA scoring function in the sidechain module. A scoring function specific RotamerConstructor is provided that comes with extensive heuristics to parametrize arbitrary compounds.
Motif finding algorithm to identify objects in 3D space, e.g. binding sites. The algorithm is based on principles of geometric hashing.
Several minor bug fixes, improvements, and speed-ups

Generalize particle scoring and rotamer construction in sidechain module. This simplifies the addition of other scoring functions in the future. Be aware of breaking changes introduced in the process! (SCWRLRotamerConstructor -> SCWRL4RotamerConstructor, changed interface of Particle and RotamerConstructor classes).
Enable possibility to use structural fragments in default modelling pipeline and build-model action
Enable possibility to enforce full coverage models including termini without template coverage in default modelling pipeline and build-model action
Modelling pipeline can track issues in the ModellingHandle object
External example scripts can now be found in extras/external_scripts
Improved support for recent compilers and libraries.
Several minor bug fixes, improvements, and speed-ups

Apply Apache Version 2.0 License to the project
2010 Dunbrack rotamer library has been replaced by an own backbone dependent rotamer library. All scripts required to reproduce the data are in extras/data_generation/rotamer_library
Penultimate rotamer library has been replaced by an own backbone independent rotamer library. All scripts required to reproduce the data are in extras/data_generation/rotamer_library
SampleMonteCarlo function moved to Python. This makes it possible to provide sampler/closer/scorer/cooler objects implemented in both, Python and C++
Action script for sidechain modelling
Allow sequence profiles as input for build-model action script.
Recipe for Docker / Singularity container
Check peptide bonds when building a RawModel. Treat as gap if bond is stereochemically problematic despite being in sequence.
Several minor bug fixes, improvements, and speed-ups

Graph optimizer has been separated from the sidechain module and can now be used for arbitrary optimization problems. Available optimization algorithms are TreePack, AStar and MonteCarlo.
Make it possible to distinguish between the scores of a loop towards a constant environment (external scores) and the scores within the loop itself (internal scores).
Most scores based on pairwise interactions are now pairwise decomposable.
Disconnected loops at several locations can be scored at once.
Avoid the usage of the DSSP executable and use the OpenStructure internal secondary structure assignment instead.
Allow to decide whether the CB atom belongs to the actual rotamer or to the constant frame in sidechain optimization. This can be useful in design questions, where the identity of a rotamer is not given and you want to allow glycine or alanine.
The naming of the entries in the StructureDB is not strictly limited to 4 letter codes anymore, arbitrary strings can be used instead.
Adding coordinates to the StructureDB does not require external tools anymore (msms, dssp etc.), internal implementations are used instead.
The data that is stored in a StructureDB can be controlled at initialization, everything except the sequence and the actual coordinates is optional.
Entries in the StructureDB can be removed again.
Allow to search positions of arbitrary copies in DynamicSpatialOrganizer by providing RT operators.
Several minor bug fixes, improvements, and speed-ups

Updated dependencies: need Eigen 3.3.0 and OST 1.7
Changes in modelling pipeline: remove clashing loop candidates early, new scores using sequence / structure profiles, new scoring weights, subrotamer optimization to reduce ring punches, and improved handling of ligands
Improved build-model action to generate oligomeric models, input multiple files, and allow for residue offsets into structure files
Improved computational runtime of relevant operations in pipeline (CCD, scorers, backbone list operations), superpositions and loop candidate clustering
High-level functionality moved consistently to modelling module
New modelling algorithms: rigid blocks to compare structures and improved fragment sampling for de novo modelling
New rotamer graph solvers for sidechains: AStar and Monte Carlo
Refactored backbone scoring to simplify extensions
Added all atom scoring with possibilities to reconstruct sidechains for selected loops, reconstruct hydrogens and minimize energy with MM
Refactored backbone list to speed it up
Added runtime profiling for developers
Several minor bug fixes, improvements, and speed-ups