v1.1.2

What's Changed

• v1.1.0 by @sokrypton in #88
• cleanup by @sokrypton in #91
• minor edits by @sokrypton in #92
• V1.1.0 by @sokrypton in #99
• V1.1.0 by @sokrypton in #100
• updating plddt and pae loss calculation to consistent with the reported plddt and pae outputs by @sokrypton in #103
• V1.1.0 by @sokrypton in #108
• V1.1.1 by @sokrypton in #110
• adding mlm support by @sokrypton in #111
• V1.1.1 beta by @sokrypton in #121
• Adding solubleMPNN weights by @KeaunAmani in #154
• Update af_cyc_design.ipynb by @stephen-rettie in #168
• added offset typing by @stephen-rettie in #169

New Contributors

• @KeaunAmani made their first contribution in #154
• @stephen-rettie made their first contribution in #168

Full Changelog: v1.0.8...v1.1.2

v1.1.0-post

Full Changelog: v1.1.0...v1.1.0-post

v1.1.0

• 15Oct2022 - v1.1.0
  • integrating proteinMPNN!
  • bugfix for sidechain loss
• 17Nov2022
  • updating pae/plddt loss calculation to be consistent with pae/plddt outputs

v1.0.9

Do not use, skip to v1.1.0!

What's Changed

• expanding callback functions by @sokrypton in #82
• refactoring the GD optimization code by @sokrypton in #83
• refactoring the optimization code to use optax library by @sokrypton in #84

Full Changelog: v1.0.8...v1.0.9

v1.0.8

• custom callback functions ([pre|loss|pos]_callback) have been refactored to be more flexible.
  • Supported input arguments include: ["inputs", "outputs", "params", "opt", "seq", "aux", "key"].
  • The pre_callback function can be used to modify inputs before prediction, loss_callback to add custom loss.
• adding support for Optax optimizers

Full Changelog: v1.0.7...v1.0.8

v1.0.7

• refactoring design.py to add design_pssm_semigreedy() protocol, which is a wrapper around design_semigreedy(seq_logits=), and can be used to input/learn PSSM for biased optimization.
• adding example peptide_binder_design.ipynb targeted for peptide binder hallucination/design.
• adding finer control over what models are used during optimization.
• fixing RAM memory leaks, clear_mem() now also does garbage collection
• fixing integration with TrDesign that got broken in v1.0.6
• WARNING: custom loss_callback has been refactored to only take (inputs, outputs) as input. inputs dictionary now contains ["params","opt","seq"]

v1.0.6

• support for alphafold-multimer model = mk_afdesign_model(..., use_multimer=True)
• support for experimentally resolved loss model.set_weights(exp_res=1)
• support for multichain design/hallucination for fixbb, hallucination and partial protocols: model.prep_inputs(..., copies=2)
• support to fix the sequence for certain positions model.prep_inputs(..., fix_pos="1-10") (supported in protocols "fixbb" and "partial")
• binder protocol improved, prior protocol would try to optimize number of contacts per target, new default is to optimize number of contacts per binder position. Number of contacts per binder position can be controlled with model.set_opt("i_con",num=1) and number of positions that should be contact with model.set_opt("i_con",num_pos=5)
• implementing David Jones'-like protocol for semi-greedy optimization, where positions are selected based on plddt, and after 20 tries, the mutation that decreasing loss the most is accepted. model.design_semigreedy()
• WARNING: the returned pLDDT is now in the "correct" direction (higher is better)

v1.0.5

• 28June2022 - v1.0.1 - Major code reorganization/refactoring to add support for callbacks (to allow integration w/ other tools during design) and to avoid clashes with existing trrosetta/alphafold installations. (eg. af → colabdesign, af.src → colabdesign.af and alphafold → colabdesign.af.alphafold).
• 05July2022 - v1.0.2 - Major code cleanup, removing duplicate code. Adding support for custom loss functions.
• 11July2022 - v1.0.3 - Improved homo-oligomeric support. RMSD and dgram losses have been refactored to automatically save aligned coordinates. Multimeric coordinates now saved with chain identifiers.
• 23July2022 - v1.0.4 - Adding support for openfold weights. To enable set mk_afdesign_model(..., use_openfold=True).
• 31July2022 - v1.0.5 - Refactoring to add support for swapping batch features without recompile. Allowing for implementation of AF2Rank!
• 7Sept2022 - re-release with a few minor bugfixes

v1.0.1

saving current version before major update

v1.0.0 - Major refactoring of the code

Adding support for partial hallucination and binder redesign and various compilation speedups.