ProtStruc
What is ProtStruc?
ProtStruc is a Python package for handling protein structures, especially for deep learning applications, through a simple, flexible, and efficient representation of protein structures.
There are many ways to represent protein structures in various deep learning applications:
Year & Source | Category | Deep learning application | Protein structure representation |
---|---|---|---|
2020, PNAS | Structure prediction | TrRoseTTA | Inter-residue geometry |
2021, Science | Structure prediction | RoseTTAFold | Inter-residue geometry |
2021, Nature | Structure prediction | AlphaFold2 | Orientation & translation of backbone frames centered at Ca's |
2022, Patterns | Structure prediction | DeepAb | Inter-residue geometry |
2023, Nat. Commun. | Antibody structure prediction | IgFold | Oritentation & translation of backbone frames centered at Ca's |
2022, arXiv | Structure generation | FoldingDiff | Three backbone dihedrals and three bond angles |
2022, NeurIPS | Structure generation | DiffAb | Orientation & translation of backbone frames centered at Ca's |
2022, Science | Inverse-folding | ProteinMPNN | k-nearest neighbor graph |
2022, arXiv | Inverse-folding | PiFold | Inter-atomic/residue distance, backbone dihedrals and bond angles, orientation of residue frame, inter-residue orientations |
2023, Science | Structure prediction | ESMFold | Orientation & translation of backbone frames centered at Ca's |
2023, ICML | Structure generation | FrameDiff | Orientation & translation of backbone frames centered at Ca's, an additional torsion angle for oxygen atom |
2022, ICML | Structure generation | Hierarchical Equivariant Refinement Network (HERN) | |
2022, arXiv | Sequence-structure co-design | Multi-channel Equivariant Attention Network (MEAN) | |
2023, ICML | Sequence-structure co-design | Dynamic multi-channel Equivariant Attention Network (dyMEAN) |