Skip to main content
. Author manuscript; available in PMC: 2024 Jan 5.
Published in final edited form as: Structure. 2022 Dec 15;31(1):111–119.e2. doi: 10.1016/j.str.2022.11.012

Figure 5. AlphaFold2 predicts peptide structures better than alternative computational methods PEPFOLD-3 (PF), OmegaFold (OF), RoseTTAFold (RF), and APPTEST (AT).

Figure 5.

A-F. Distributions of average Cα RMSD per residue between the secondary structural regions of the NMR ensemble compared to each alternative computational prediction method. Peptide groups (A) α-helical membrane protein (AH MP), (B) α-helical soluble (AH SL), (C) β-hairpin peptides (BHPIN), (D) disulfide rich peptides (DSRP), (E) mixed membrane peptide (MIX MP), and (F) mixed soluble peptide (MIX SL) are shown. AlphaFold2 performed statistically significantly better than PEPFOLD3, OmegaFold, RoseTTFold and APPTEST, in all peptide classes except for MIX MP and MIX SL. For MIX MP, all methods perform with similar accuracy except for APPTEST. Interestingly, OmegaFold performed just as well as AF2 for MIX SL, and both outperformed the other methods. A paired two-side student t-test was used to calculate the statistical difference of each distribution with respect to AF2 and p value depicted by * < 0.05, ** < 0.01, *** < 0.001, and **** < 0.0001.