Table 2.
Comparison of full-length-level modeling results by AlphaFold2-based methods and protein language model (PLM)-based methods on the 65 CASP14 full-length targets. Here, AlphaFold2-Single is the default AlphaFold2 pipeline, with the only query sequence as the input MSA. p-values were calculated between TM-scores by AlphaFold2 and others using paired one-sided Student’s t-tests. #{TM > 0.5} is the number of targets with a TM-score > 0.5.
| Method | Method Type | TM-Score | p-Value | #{TM > 0.5} |
|---|---|---|---|---|
| AlphaFold2 | AlphaFold2-based | 0.8514 | - | 60 |
| ColabFold | 0.8461 | 2.37 × 10−1 | 61 | |
| OpenFold | 0.8375 | 1.46 × 10−1 | 59 | |
| Uni-Fold | 0.8561 | 9.82 × 10−1 | 61 | |
| AlphaFold2-Single | PLM-based | 0.5164 | 6.88 × 10−12 | 30 |
| ESMFold | 0.6676 | 1.80 × 10−10 | 41 | |
| OmegaFold | 0.6728 | 4.42 × 10−6 | 43 |