Incorporation of atomistic molecular dynamics simulations - Proteogram v2

What's Changed

• Molecular dynamics simulations for new energy calculations and structure refinement - Proteogram v2
• Reorg repo for v1 and v2 approaches
• Dealing better with PTMs
• Better evaluation logic
• Inclusion of Foldseek as a benchmark method
• Single chain inference, update licenses and improved documentation

Full Changelog: v0.0.2...v0.0.3

Attached to this release

• Demo model file for testing - proteogram_demo_resnet18_finetuned_lr0.001_bs8_e29_85.5acc.pt
• Demo embedding vectors for all proteins used in the v2 study (2724 proteogram embeddings) - proteogram_demo_embeddings_scope2.08-nr60_20-200.pkl (Note: these are a limited set of embeddings to be able to demo the scripts only, not for production workflows) - a limited, non-redundant (60% sequence identity cutoff) set of Proteogram embeddings for proteins 20-200 residues long (set that was used for training and eval in the e2e pipeline)

Acknowledgements

• Thank you to @charms-research for advising on the molecular dynamics simulation application for this release
• Thank you to @swapnilashtekar for contributing a path for reproducibility and validating the codebase and workflows

ISCB Paper Submission Release

The primary purpose of this release is to add a model file for measuring similarity. Place the name of this model file with its path into your config.yml, e.g., model_file: /your/path/resnet18_scratch_without_sampler_bs8_lr0.01.pt.