Tscode

Discontinued OpenBabel FF support
- Updated prune_conformers_rmsd group criteria, avoiding group numbers that have (on average) less than 20 active structures.
- Disconnected the Openbabel Force Field calculator from the embedder, as the XTB implementation proved uniquely versatile and robust for more sophisticated manipulations. The interface is still present in the code (inside calculators) for reference, external utility purposes and potential future re-adoption for specific tasks. Updated the rest of the documentation accordingly.

RMSD pruning significative speedup, increased extent of parallelization throughout, keywords priority
- Small bug fixes.
- Added priority to keywords, that dictate their order of execution. Options are set first (priority 1), then priority 2 keywords that modify attributes and depend on priority 1 keywords (for now, just DIST).
- Added a status_dump call after generate_candidates (debug keyword).
- Changed spring stiffness for constrained optimizations to dynamic values (0.25 to 1 for FF, 1 to 2 for SE, in Eh/bohrs). Less scrambling, more accurate poses.
- Multiembeds now are run through self.avail_cpus max_workers instead of half.
- prune_conformers_rmsd is now completely compiled with numba, cached, parallelized and ~30 times faster. The old similarity-graph-based logic was also removed in favor of the removal of any structure at the first instance of a similar one. In the future, all pruning functions should work similarly.
- Cyclical embed now generates fewer candidates, as it discards them if they are too similar to others that share the same pivots and conformation ids (RMSD-based).
- Adjusted default threshold for RMSD similarity from 0.25 back to 0.5 A (benchmarked to retain all methylcyclohexane conformers with the new pruning algorithm).
- Reduced checkpoint dump frequency (embedder.options.checkpoint_frequency) from 20 to 50, as writing large (>20k) structures so often can slow down execution and clutter the logfile.
- Reversed order of this CHANGELOG.md file, to display the most recent updates on top.

Memory usage, pruning refinements, multiembed improvements, molecule attributes
- Significatively reduced memory usage of the prune_conformers functions, avoiding using the wasteful similarity_mat for a lighter, faster set().
- Reduced default values for ROTRANGE and STEPS for cyclical embeds (from 90, 9 to 45, 5).
- TFD similarity refining is now only carried out for single molecules, not for multimolecular embeds.
- Added SIMPLEORBITALS keyword. All orbitals will be of Single type. Reduces the number of pivots for each molecule, and consequently the number of candidates that will be generated.
- Added generation of orbitals for main embedder in multiembed runs. Aids debugging and allows for a better bird's-eye view in the initial printout.
- Input file: added custom attributes in molecule line (mtd> mol.xyz 2A 3A 7x charge=-1). Any attribute is set directly on the Hypermolecule class, and its declaration is recorded in the logfile. For now, the only active attribute is 'charge' and it is passed to the mtd> operator. The approach should be easily applicable to set future molecule-specific settings.
- Fixed SHRINK printout bug when not specifying embed distances.
- Multiembed child embedders are passed more options now: shrink (with shrink_multiplier) and simpleorbitals

**CHANGED CONSTRAINTS SPECIFICATION (NON-BACKWARDS COMPATIBLE)**
RunEmbedding refactoring, stability improvements, internal cleanup of old code
"Dihedral embeds" are now part of scan>
- Various small fixes for stability purposes and printout beautification.
- write_structures can now also align ensembles based on the moments of inertia (align='moi')
- Renamed output of refine> runs 'ensemble' instead of 'poses', which is now only for when an embed is carried out.
- Cleaned some unused junk code from the past (fast_score, hyperneb, opt_iscans, TS keyword, and other experiments)
- Dihedral "embeds" are now just a part of the scan> operator, as no real embed was carried out.
- Added printout of input file after the banner in the logfile for easier tracebacks.
- Renamed mtd_csearch> to mtd_search> (solves a bug where csearch> was called instead).
- Moved RunEmbedding back to embedder.py, and cleaned up the class inheritance. This allowed less redundant and tidier code at the expense of having a larger embedder.py file (~2300 lines).
- Expanded dump_status to reflect the different constraints at different steps of optimization (all constraints or just fixed). Also added the target distance for each in the printout.
- **CHANGED CONSTRAINTS SPECIFICATION (NON-BACKWARDS COMPATIBLE)** - now fixed constraints are specified with UPPERCASE letters instead of a/b/c, and interaction constraints are specified with lowercase letters instead of x/y/z. Expands the number of each type of constraint that is possible to specify (from 3 to 26).

CONCURRENT OPTIMIZATIONS AND EMBEDDINGS!
Crest metadynamic conformational searches with mtd>
- xtb_opt now reads output geometry from the trajectory file rather than from xtbopt.xyz. This was needed for multiprocessing capability as it is not possible to override the default name for the optimized geometry (xtbopt.xyz) from the input section.
- Replaced exit() calls with sys.exit()
- xtb_opt now creates a new folder and works inside of it, making it easier to transition into a multiprocessing workflow.
- xtb_pre_opt passes bond constraints to xtb_opt through a string now (constrain_string) instead of a file.
- Implemented parallel (multiprocess) optimization for force_field_refining (2 cores/thread, as many threads to use all CPUs) and optimization_refining (4 cores/thread, as many threads to use all CPUs). Embedder.threads variable currently not in use.
- Added CHARGE keyword (specify charge to be passed to calculator).
- Removed VMD printout for simple runs (write_vmd) but kept anchor debugging (write_anchor_vmd).
- Added metadynamic conformational search through CREST (mtd_csearch>, or mtd>).
- Changed the way xtb_opt deals with distance constraints, now using very stiff springs instead of exact fixing (better when more than one distance constraint is specified).
- Reinstated fitness_refining after every optimization step, which is now based on a cumulative deviation from the imposed pairing distances.

- Expanded debug functionality (DEBUG keyword): printouts of RunEmbedding status (dump_status)
- Added a xtb_pre_opt wrapper for xtb_opt that retains every bond present in the initial set of graphs, in addition to provided constraints. This will greatly increase the quality of structures coming from embeds into force_field_refining. Added before force_field_opt for embed runs (>1 self.objects) that use XTB as FF calculator.
- Introduced orbital subtypes (atom.subtype) to offer more flexibility for embed types. These do not alter default pairing_dists but allow for specific orbital arrangements around the reactive atom.
- New Ketone atom subtype: 'trilobe' - three lobes at the opposite end of alkoxide/sulfonamide substituents relative to oxygen.
- Added the DRYRUN keyword. Skips lengthy operations (operators, embedding, refining) but retains other functions and printouts. Useful for debugging and checking purposes.
- Compacted output information: individual candidate details only printed with DEBUG keyword, average time per structure and estimate of completion given instead every 20 iterations when saving checkpoints