Table 1.
Scores used in protein 3D structure comparison and evaluation. Distance-based and contact-based similarity scores are used in experimental evaluation of homology models. Other scores are used for quality check such as physics-based, knowledge-based and combined scores.
| Score | Description | Reference |
|---|---|---|
| Distance-based similarity scores: | ||
| RMSD | Root mean square deviations | |
| wRMSD | weighted RMSD | |
| RMS of dihedral angles | Root mean square of dihedral angles | [42] |
| GDT | Global distance test employing local global alignment (LGA) program | [43] |
| GDT_TS | GDT total score: Iterations superposing sets of 3, 5 and 7 consecutive Cα atoms (thresholds 1, 2, 4, and 8 Å) | [44] |
| GDT_TL/GDT_HA | GDT high accuracy scores: Finer thresholds than GDT_TS (0.25, 0.5, 1, and 2 Å) | [14], [44] |
| TM-score | Variations between Cα atoms weighting residues at shorter distances | [45] |
| TM-align | Based on TM-score for evaluating global variations | [45] |
| MaxSub | Normalized score from large subset of Cα atoms | [46] |
| SphereGrinder | Specialized for large predicted models | [47] |
| Contact-based similarity scores: | ||
| CAD | Contact area difference | [48] |
| CAD-score | Contact area difference | [49] |
| Physics-based quality scores: | ||
| Molprobity score | Global (whole protein) and local (small regions) perspectives | [50] |
| What IF | Surface area, solvent accessibility, and hydrophobicity checks | [51] |
| PROCHECK | PROgram to CHECK stereochemical quality | [52] |
| Knowledge-based quality scores: | ||
| QMEAN | Qualitative Model Energy ANalysis | [53] |
| DOPE | Discrete Optimized Protein Energy | [54] |
| PROSAII | PROtein Structure Analysis II | [55] |
| Combined quality scores: | ||
| MetaMQAP | Meta-methods for quality assessment of protein models | [56] |
| Machine learning methods | Using support vector machine (SVM) to combine scores | [57] |
| Z-score | Any arbitrary score function based on sum of a number of scores (e.g. force field energies, GDT, etc.) May or may not be normalized |
|
| Other methods: | ||
| Experimental studies | Validation can be done by other experimental data, such as molecular dynamics simulations, spectroscopic methods, binding analysis (e.g. calculations of dissociation/inhibition or Kd/Ki constants) | [58], [59] |