Table 2.
Methods evaluated for amino acid chirality in the predicted structures
| Methods | PDBfixeda X-ray structures used as input | OpenMM used | Average number of chiral amino acids per modelb | Number of D-amino acids introduced in the MHC | Number of D-amino acids introduced in the peptide |
|---|---|---|---|---|---|
| MHCfold with MODELLER | 174 | 1 | 0 | ||
| AlphaFold2-FineTune | ✔ | 173 | 0 | 0 | |
| APE-Gen 2.0 | ✔ | 263 | 22 | 12 | |
| PANDORA | 263 | 0 | 0 | ||
| SwiftMHC | ✔ | 174 | 0 | 0 | |
| SwiftMHC-OpenMM | ✔ | ✔ | 174 | 9 | 0 |
a
OpenMM PDBfixer29 was used to restore missing side-chain atoms in some of the 202 X-ray structures, which resulted in the introduction of 11 D-form amino acids.
b
Some methods model the entire class I MHC protein structure, while others model only the G-domain, leading to differences in the total number of amino acids.