2.0.1

• a fix to conformer generation including R atom during position generation (#75)

2.0.0

The major release 2.0.0 contains two major features: Active Learning and Parallelisation Acceleration.

The best way to familiarise yourself is going to be via using the Jupiter notebooks. Please get in touch if you spot any errors, or if you'd like to see any improvements.

Active learning is carried out with scipy and supports multiple models and querying strategies. This is further parallelised with Dask, which can split the work across CPUs, different PCs (over -ssh), cluster (PBS, SLURM, etc), and more.

The major changes have been encoded in the new class called ChemSpace. However, some of the previous minute functions were modified to simplify the interactions between Dask and them. This means that not all previous functions have the same interface.

This release also contains a number of small fixes.

What's Changed

• Daskify rlist by @bieniekmateusz in #69

Minor

• Libraries initiated on request by @bieniekmateusz in #34
• Single SDFs for libraries by @bieniekmateusz in #37
• library order by @bieniekmateusz in #38
• Gnina error handling by @bieniekmateusz in #39
• Wording: Wrong Atoms picking for building a molecule by @bieniekmateusz in #40
• ANI - exploded conformers by @bieniekmateusz in #42
• A scaffold: allow the use of a previously minimised structure as a starting scaffold by default. by @bieniekmateusz in #43
• Builder: growing not just from hydrogens by @bieniekmateusz in #44
• We don't need the hydrogens. by @bieniekmateusz in #45
• Make more use of openmm pdbfile by @bieniekmateusz in #46
• A growing vector splits the molecule into several components by @bieniekmateusz in #47
• Saving MDAnalysis-readable .pdb files by @bieniekmateusz in #49
• Keeping original scaffold by @bieniekmateusz in #52
• update HBD/HBA calc for Lipinski by @RPirie96 in #51
• No Hs in RMS to avoid the wrong hydrogens being compared by @bieniekmateusz in #55
• Present gnina "CNNAffinity->IC50" as Kd [nM] with pint-pandas by @bieniekmateusz in #59
• Linkers and RGroups are now dataframes - easier API access by @bieniekmateusz in #62
• CLI high throughput scoring by @jthorton in #64

Full Changelog: v1.3.0...v2.0.0

1.3.0

• Compatibility with openff-toolkit >=0.11 (thanks to @jeeberhardt for #29 #30)

1.2.1

• library dataset has better coordinates (Ben)
• visualisations with pandas work better
• error handling

1.2.0

• Adding the ability to work with linkers in addition to R-groups,
• Embedding a linker library (see and cite https://www.sciencedirect.com/science/article/abs/pii/S0968089623000421)

See the log for more.

1.0.2

• conda ready release

1.0.1

Updates:

• embeds the openmm-ml plugin
• optimisation now allows you to specify whether to use the cpu or gpu (which also fixes the bug due to both being deployed as the same time)

1.0.0

An Open-Source Molecular Builder and Free Energy Preparation Workflow by Mateusz K. Bieniek, Ben Cree, Rachael Pirie, Joshua T. Horton, Natalie J. Tatum, and Daniel J. Cole.

Further Information
Please see cole-group.github.io/fegrow for full installation instructions, documentation and acknowledgements.

And see the tutorial to get started.