AlphaFold2 examples from CASP 14: Difference between revisions

Revision as of 00:18, 4 March 2021

This page is under construction. Eric Martz 01:03, 22 February 2021 (UTC)

Prediction of protein structures from amino acid sequences, homology modeling, has been extremely challenging. In 2020, breakthrough success was achieved by AlphaFold2^[1], a project of DeepMind. For an overview of this breakthrough, documented by the bi-annual prediction competition empirical models, please see 7jtl. Below are illustrated some examples of predictions from that competition.

SARS-CoV-2 ORF8

Following the discussion by Rubiera^[1], our first example will be SARS-CoV-2 protein ORF8, a protein that contributes to virulence in COVID-19^[2]. CASP 14 classified ORF8 as a "free modeling" (FM) target^[3], meaning that there were no adequate empirical templates for homology modeling. This was easily confirmed. When the amino acid sequence of ORF8 is submitted to Swiss Model, it reports the best templates for homology modeling. When the two empirical models that were not available during CASP 14 are excluded (7jtl and 7jx6), the best template offered, chain B of 3afc, covers only 36% of the length of ORF8 at 13.2% sequence identity, with a 4-residue untemplated gap in the sequence alignment. This template would not be adequate for constructing a useful model.

X-Ray Structures for ORF8

The quality of predictions for the structure of ORF8 are judged by comparison with X-ray crystallographic empirical models which were not available to the groups making predictions. Shortly after the CASP 14 competition (summer 2020), two X-ray crystal structures were reported for ORF8: 7jtl released August 26, 2020, and 7jx6, released September 23, 2020. The resolutions are 2.0 and 1.6 Å respectively, and both have worse than average Rfree values.

Click the green links below to change the molecular scene. Drag to rotate. Zoom the molecule with your mouse wheel, or Shift-Drag up/down.

Here is one chain of ORF8 from the higher resolution X-ray structure, 7jx6. These chains form disulfide-linked dimers, and the dimers form higher order multimers^[2] (not shown). Notice that the amino and carboxy ends of the chain come together to form two parallel beta strands of a beta sheet. Also notice that there are 3 disulfide bonds. An accurate prediction would include both of these features.

The two X-ray structures agree very well^[4]. The only substantial disagreement is for a large surface loop, sequence range 48-57. See the Table I below for RMSD values.

ORF8 is not a novel fold

Less than 2% of new empirically-determined structures have novel folds; that is, folds not aready represented in the PDB^[5]. When chain A of 7jx6 was submitted to Dali^[6] (February, 2021), the top hit was the N-terminal domain of the two domains in 5a2f, the CD166 human cell surface receptor involved in activation of T lymphocytes. The Z-score was 7.1, and 88 alpha carbons superposed with RMSD 3.2 Å. Swiss-PdbViewer obtained RMSD 1.95 Å for 48 alpha carbons^[7]. Dali reported the identity as 6% in its structure-based sequence alignment. Sequence alignment by MAFFT^[8] obtained 18% sequence identity using more and larger gaps. The structural similarity between Dali's top hit and 7jx6^[9] is not as close as for AlphaFold2's prediction, but is closer than the 2nd best prediction (see Table I below). In conclusion, ORF8 does not have a novel fold^[10].

AlphaFold2 Prediction for ORF8

The quality of a prediction in CASP is judged, in large part, by the Global Distance Test Total Score, GDT_TS. AlphaFold2's predicted structure^[11] has a GDT_TS score of 87. (A score of 0 is meaningless, and a score of 100 means perfect agreement with an X-ray crystal structure.) 87 means the model is close to the accuracy of an X-ray crystal structure^[4]. The structure predicted by AlphaFold2 is almost as close to the X-ray crystallographic model 7jx6 as is the independently-determined X-ray structure 7jtl. AlphaFold2 predicted the positions of 92 amino acids. (CASP 14 excluded residues 48-59, a 12-residue surface loop, from the target residues^[3].) See Table I below for RMSD values. The prediction was largely accurate regarding salt bridges and cation-pi interactions (see Tables II and III below).

Table I. ORF8 Predictions Superposed With Chain A of 7jx6
Model	GDT_TS	Disulfde Bonds	Cα RMSD, Å	Cα Superposed	RMSD Including Sidechains, Å	Atoms Superposed
7jtl:A	88^[12]	3	4.02 0.66	102/102 (100%) 87/102 (85%)	4.3 1.58	829/829 (100%) 709/829 (86%)
AlphaFold2	87	3	2.58 1.25	92/92 (100%) *83/92 (90%)**	3.23 1.91	747/748 (100%) 679/748 (91%)
Dali top hit^[13] 5a2f	53^[12]	na	3.2 1.95	92/92 (100%) 48/92 (52%)	na	na
2nd Best*	43	0	5.33 1.71	92/92 (100%) 38/92 (41%)	6.54 5.86	747/748 (100%) 324/748 (43%)
3rd Best§	33	0	13.37 †	92/92 (100%) †	14.50 †	747/748 (100%) †
Zhang-TBM Server	27	0	14.90 †	92/92 (100%) †	15.61 †	747/748 (100%) †
Rosetta Server	26	(2‡)	14.99 †	92/92 (100%) †	16.07 †	747/748 (100%) †

Superpositions by "Magic Fit"^[14] of Swiss-PdbViewer 4.1.

Superpositions by "Iterative Magic Fit"^[4] of Swiss-PdbViewer 4.1.

na: Not Applicable.

*Second best: Group of Xian Ming Pan, Tsinghua University, Beijing.

§Third best: Group of Alberto Perez, University of Florida, Gainsville.

† Iterative Magic Fit was unable to superpose.

‡ Neither disulfide bond is correct.

Second Best Prediction for ORF8

In CASP 14, 70 research groups and 42 automated servers predicted structures for ORF8. The median GDT_TS score for all 112 predictions was 26. AlphaFold2 made the best prediction (GDT_TS 87). The second best prediction was by the group of Xian Ming Pan^[4], with GDT_TS 43 (see Table I above). The fold and topology were predicted correctly, but the details are far less accurate than those in AlphaFold2's prediction. The 2nd best prediction has no disulfide bonds. This prediction was largely incorrect regarding salt bridges and cation-pi interactions (see Tables II and III below).

Third Best Prediction for ORF8

The third best prediction for ORF8 was by the Perez Lab, with GDT_TS 33 (see Table I above). It correctly predicted the parallel beta strands formed by the amino and carboxy terminal ends of the chain. When the 2-stranded parallel beta strands formed by the ends of the chains are superposed, the remainder superposes poorly^[15]. This prediction has no disulfide bonds. The salt bridge Arg86:Asp98 is correctly predicted, along with two incorrectly predicted salt bridges.

Top Prediction by an Automated Server

Among predictions by automated servers for all ~100 CASP 14 targets, the top ranking server was QUARK from the Yang Zhang group (Univ. Michigan). For ORF8, the Zhang-TBM server made the best server prediction with a GDT_TS of 27. (The prediction by QUARK was almost as good, GDT_TS 26.) The prediction has the two chain termini not parallel, and the amino terminus is not a beta strand, differing in both respects from the X-ray model. Also, no disulfide bonds are predicted. The salt bridge Arg86:Asp98 is correctly predicted, along with several incorrectly predicted salt bridges. The structural superposition is very poor and is not shown.

Baker Rosetta Server Prediction for ORF8

Among predictions for all ~100 CASP 14 targets, the group of David Baker ranked second. The Rosetta Server of the Baker group ranked 18th overall, but was the 4th ranked server^[16]. For ORF8, the Rosetta Server prediction GDT_TS was 26, a bit better than the median of 23. The Rosetta Server's prediction for ORF8 has the two termini far apart (Cα 13 Å or farther apart), a substantial difference from the X-ray structure (Cα mostly ~5 Å apart). It predicts two disulfide bonds, but neither matches the pairs of Cys residues in the actual disulfide bonds. The salt bridge Arg86:Asp98 is correctly predicted, along with one incorrectly predicted salt bridge. The structural superposition is very poor and is not shown.

↑ CASP14: what Google DeepMind’s AlphaFold 2 really achieved, and what it means for protein folding, biology and bioinformatics, a blog post by Carlos Outeir al Rubiera, December 3, 2020.
↑ ^2.0 ^2.1 Flower TG, Buffalo CZ, Hooy RM, Allaire M, Ren X, Hurley JH. Structure of SARS-CoV-2 ORF8, a rapidly evolving immune evasion protein. Proc Natl Acad Sci U S A. 2021 Jan 12;118(2). pii: 2021785118. doi:, 10.1073/pnas.2021785118. PMID:33361333 doi:https://dx.doi.org/10.1073/pnas.2021785118
↑ ^3.0 ^3.1 Summary and Classifications of Domains for CASP 14.
↑ ^4.0 ^4.1 ^4.2 ^4.3 Superposition by Swiss-PdbViewer's iterative magic fit. This starts with a sequence alignment-guided structural superposition, and then superposes subsets of the structures to minimize the RMSD. Eight intermediate structures were generated by the Theis Morph Server by linear interpolation.
↑ Cuff AL, Sillitoe I, Lewis T, Clegg AB, Rentzsch R, Furnham N, Pellegrini-Calace M, Jones D, Thornton J, Orengo CA. Extending CATH: increasing coverage of the protein structure universe and linking structure with function. Nucleic Acids Res. 2011 Jan;39(Database issue):D420-6. doi: 10.1093/nar/gkq1001. , Epub 2010 Nov 19. PMID:21097779 doi:https://dx.doi.org/10.1093/nar/gkq1001
↑ Holm L. DALI and the persistence of protein shape. Protein Sci. 2020 Jan;29(1):128-140. doi: 10.1002/pro.3749. Epub 2019 Nov 5. PMID:31606894 doi:https://dx.doi.org/10.1002/pro.3749
↑ Using Swiss-PdbViewer's Fit from Selection with 102 residues selected from each structure, followed by Improve Fit.
↑ Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013 Apr;30(4):772-80. doi: 10.1093/molbev/mst010. Epub 2013 Jan, 16. PMID:23329690 doi:https://dx.doi.org/10.1093/molbev/mst010
↑ Structural superposition by Dali. Interpolation by the Yale Morph2 Server. Homogenization method: homology modeling. No minimization. This produced a 9-model file where model 1 was 7jx6, and models 2-9 were interpolations. 5a2f residues 28-133 were added as model 10 (black in the molecular scene).
↑ The interpretation of Dali's result to mean that ORF8 does not have a novel fold was kindly confirmed by Liisa Holm, personal communication to Eric Martz, February, 2021.
↑ Download AlphaFold2's predicted structure for ORF8 from T1064TS427_1-D1.pdb.
↑ ^12.0 ^12.1 See #GDT_TS Calculations.
↑ See #ORF8 is not a novel fold.
↑ Superposition by Swiss-PdbViewer's magic fit. This is a sequence alignment-guided structural superposition. Eight intermediate structures were generated by the Theis Morph Server by linear interpolation.
↑ Superposition by Swiss-PdbViewer's Explore Fragment Alternate Fits, which does not use sequence information. Eight intermediate structures were generated by the Theis Morph Server by linear interpolation.
↑ For all targets in CASP 14, the top two servers were QUARK and Zhang-server (which were not significantly different at a Z-score sum of 62.9), followed by Zhang-CEthreader (55.9) and BAKER-ROSETTASERVER (55.3).

ORF8 Sidechain Accuracy

AlphaFold2's predictions for sidechain positions seem fairly good, while sidechain positions in the 2nd best prediction seem poor. This conclusion is based on three types of observations:

Table I gives RMSD values for all atoms, which is one indication of sidechain accuracy.
Prediction of 7jx6 and 3afc.
Visualization of the distributions of charges on the surfaces.

Salt Bridges and Cation-Pi Interactions

AlphaFold2's prediction was correct for 4/5 interactions, with one incorrect interaction.
- AlphaFold2's prediction was correct for one of two salt bridges, and predicted no incorrect salt bridges.
- AlphaFold2's prediction was correct for three of three cation-pi interactions, but predicted one incorrect interaction.
The 2nd best prediction was correct for 1/5 interactions, with 2 incorrect interactions.
- The 2nd best prediction was correct for one of two salt bridges, but predicted two incorrect salt bridges.
- The 2nd best prediction failed to predict any of the three cation-pi interactions, predicting zero interactions.

Table II. Salt Bridge Prediction Accuracy
7JX6	7JTL	AlphaFold2	2nd Best
R101:D112 (AB)	R101:D113 (AB)	R86:D98	R86:D98
R115:D119 (AB)	R115:D119 (AB)	–	R100:E4
K44:E59 (AB)	K44:E59 (AB)	K29:E44	–
–	–	–	K78:E77

Bridges in the same row are identical (except for red residues). Subtract 15 from the sequence numbers in the X-ray structures for the equivalent sequence numbers in the predictions.
Black: Shortest sidechain nitrogen to sidechain oxygen distance ≤4.0 Å.
Gray: Shortest sidechain nitrogen to sidechain oxygen distance 4.4 to 4.8 Å.
–: Shortest sidechain nitrogen to sidechain oxygen distance 6 to 16 Å.
(AB): The two chains in each X-ray model.
Italics: erroneous prediction.

Table III. Cation-Pi Prediction Accuracy
7JX6	7JTL	AlphaFold2	2nd Best
R101:Y46+Y108 (AB)	R101:Y46+Y108 (AB)	R86:Y31+Y96	–
K44:F108 (B)	K44:F108 (AB)	K29:F93	–
–	–	K79:F105	–

All interactions listed are deemed energetically significant by the CaPTURE Server.
Interactions in the same row are identical. Subtract 15 from the sequence numbers in the X-ray structures for the equivalent sequence numbers in the predictions.
Italics: erroneous prediction.
The 2nd best prediction has no cation-pi interactions.
(AB): The two chains in each X-ray model.

Visualization of Surface Charge Distributions

GDT_TS Calculations

GDT_TS values for predictions are taken from CASP 14 results. GDT_TS values for 7JTL and 5A2F vs. 7JX6 chain A were calculated using the AS2TS server of Adam Zemla^[2]. See instructions for empirical models. CASP 14 reported GDT_TS 86.96 for the AlphaFold2 prediction, while the AS2TS server calculated GDT_TS 86.41.

References

↑ Senior AW, Evans R, Jumper J, Kirkpatrick J, Sifre L, Green T, Qin C, Zidek A, Nelson AWR, Bridgland A, Penedones H, Petersen S, Simonyan K, Crossan S, Kohli P, Jones DT, Silver D, Kavukcuoglu K, Hassabis D. Improved protein structure prediction using potentials from deep learning. Nature. 2020 Jan;577(7792):706-710. doi: 10.1038/s41586-019-1923-7. Epub 2020 Jan, 15. PMID:31942072 doi:https://dx.doi.org/10.1038/s41586-019-1923-7
↑ Zemla A. LGA: A method for finding 3D similarities in protein structures. Nucleic Acids Res. 2003 Jul 1;31(13):3370-4. doi: 10.1093/nar/gkg571. PMID:12824330 doi:https://dx.doi.org/10.1093/nar/gkg571

Proteopedia Page Contributors and Editors (what is this?)

Eric Martz

[rubiera-1] CASP14: what Google DeepMind’s AlphaFold 2 really achieved, and what it means for protein folding, biology and bioinformatics, a blog post by Carlos Outeir al Rubiera, December 3, 2020.

[7jtl-2] 2.0 ^2.1 Flower TG, Buffalo CZ, Hooy RM, Allaire M, Ren X, Hurley JH. Structure of SARS-CoV-2 ORF8, a rapidly evolving immune evasion protein. Proc Natl Acad Sci U S A. 2021 Jan 12;118(2). pii: 2021785118. doi:, 10.1073/pnas.2021785118. PMID:33361333 doi:https://dx.doi.org/10.1073/pnas.2021785118

[casp14domains-3] 3.0 ^3.1 Summary and Classifications of Domains for CASP 14.

[imf-4] 4.0 ^4.1 ^4.2 ^4.3 Superposition by Swiss-PdbViewer's iterative magic fit. This starts with a sequence alignment-guided structural superposition, and then superposes subsets of the structures to minimize the RMSD. Eight intermediate structures were generated by the Theis Morph Server by linear interpolation.

[cath2011-5] Cuff AL, Sillitoe I, Lewis T, Clegg AB, Rentzsch R, Furnham N, Pellegrini-Calace M, Jones D, Thornton J, Orengo CA. Extending CATH: increasing coverage of the protein structure universe and linking structure with function. Nucleic Acids Res. 2011 Jan;39(Database issue):D420-6. doi: 10.1093/nar/gkq1001. , Epub 2010 Nov 19. PMID:21097779 doi:https://dx.doi.org/10.1093/nar/gkq1001

[dali2020-6] Holm L. DALI and the persistence of protein shape. Protein Sci. 2020 Jan;29(1):128-140. doi: 10.1002/pro.3749. Epub 2019 Nov 5. PMID:31606894 doi:https://dx.doi.org/10.1002/pro.3749

[fitselimprov-7] Using Swiss-PdbViewer's Fit from Selection with 102 residues selected from each structure, followed by Improve Fit.

[mafft-8] Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013 Apr;30(4):772-80. doi: 10.1093/molbev/mst010. Epub 2013 Jan, 16. PMID:23329690 doi:https://dx.doi.org/10.1093/molbev/mst010

[yale-9] Structural superposition by Dali. Interpolation by the Yale Morph2 Server. Homogenization method: homology modeling. No minimization. This produced a 9-model file where model 1 was 7jx6, and models 2-9 were interpolations. 5a2f residues 28-133 were added as model 10 (black in the molecular scene).

[holm-10] The interpretation of Dali's result to mean that ORF8 does not have a novel fold was kindly confirmed by Liisa Holm, personal communication to Eric Martz, February, 2021.

[11] Download AlphaFold2's predicted structure for ORF8 from T1064TS427_1-D1.pdb.

[gdt_ts-12] 12.0 ^12.1 See #GDT_TS Calculations.

[nnf-13] See #ORF8 is not a novel fold.

[mf-14] Superposition by Swiss-PdbViewer's magic fit. This is a sequence alignment-guided structural superposition. Eight intermediate structures were generated by the Theis Morph Server by linear interpolation.

[fragfit-15] Superposition by Swiss-PdbViewer's Explore Fragment Alternate Fits, which does not use sequence information. Eight intermediate structures were generated by the Theis Morph Server by linear interpolation.

[serverranks-16] For all targets in CASP 14, the top two servers were QUARK and Zhang-server (which were not significantly different at a Z-score sum of 62.9), followed by Zhang-CEthreader (55.9) and BAKER-ROSETTASERVER (55.3).

[af2-1] Senior AW, Evans R, Jumper J, Kirkpatrick J, Sifre L, Green T, Qin C, Zidek A, Nelson AWR, Bridgland A, Penedones H, Petersen S, Simonyan K, Crossan S, Kohli P, Jones DT, Silver D, Kavukcuoglu K, Hassabis D. Improved protein structure prediction using potentials from deep learning. Nature. 2020 Jan;577(7792):706-710. doi: 10.1038/s41586-019-1923-7. Epub 2020 Jan, 15. PMID:31942072 doi:https://dx.doi.org/10.1038/s41586-019-1923-7

[zemla-2] Zemla A. LGA: A method for finding 3D similarities in protein structures. Nucleic Acids Res. 2003 Jul 1;31(13):3370-4. doi: 10.1093/nar/gkg571. PMID:12824330 doi:https://dx.doi.org/10.1093/nar/gkg571

[1]

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[2]

@@ Line 18: / Line 18: @@
 ===ORF8 is not a novel fold===
-Less than 2% of new [[empirically-determined structures]] have novel folds; that is, folds not aready represented in the [[PDB]]<ref name="cath2011">PMID: 21097779</ref>. When chain A of [[7jx6]] was submitted to Dali<ref name="dali2020">PMID: 31606894</ref> (February, 2021), the top hit was the N-terminal domain of the two domains in [[5a2f]], the CD166 human cell surface receptor involved in activation of T lymphocytes. The Z-score was 7.1, and 88 alpha carbons superposed with RMSD 3.2 Å. Swiss-PdbViewer obtained RMSD 1.95 Å for 48 alpha carbons<ref name="fitselimprov">Using Swiss-PdbViewer's ''Fit from Selection'' with 102 residues selected from each structure, followed by ''Improve Fit''.</ref>. Dali reported the identity as 6% in its structure-based sequence alignment. Sequence alignment by MAFFT<ref name="mafft">PMID: 23329690</ref> obtained 18% sequence identity using more and larger gaps. <scene name='87/875686/Dali_5a2f_vs_7jx6_yale/2'>The structural similarity between Dali's top hit and 7jx6</scene><ref name="yale">Structural superposition by Dali. Interpolation by the [http://www2.molmovdb.org/wiki/info/index.php/Morph2_Server Yale Morph2 Server]. Homogenization method: homology modeling. No minimization. This produced a 9-model file where model 1 was 7jx6, and models 2-9 were interpolations. 5a2f residues 28-133 were added as model 10 (black in the molecular scene).</ref> is not as close as for AlphaFold2's prediction, but is closer than the 2nd best prediction (see Table I below). Dali's top hit has a single disulfide bond (compare with Table I). In conclusion, '''ORF8 does not have a novel fold'''<ref name="holm">The interpretation of Dali's result to mean that ORF8 does not have a novel fold was kindly confirmed by Liisa Holm, personal communication to [[User:Eric Martz|Eric Martz]].</ref>.
+Less than 2% of new [[empirically-determined structures]] have novel folds; that is, folds not aready represented in the [[PDB]]<ref name="cath2011">PMID: 21097779</ref>. When chain A of [[7jx6]] was submitted to Dali<ref name="dali2020">PMID: 31606894</ref> (February, 2021), the top hit was the N-terminal domain of the two domains in [[5a2f]], the CD166 human cell surface receptor involved in activation of T lymphocytes. The Z-score was 7.1, and 88 alpha carbons superposed with RMSD 3.2 Å. Swiss-PdbViewer obtained RMSD 1.95 Å for 48 alpha carbons<ref name="fitselimprov">Using Swiss-PdbViewer's ''Fit from Selection'' with 102 residues selected from each structure, followed by ''Improve Fit''.</ref>. Dali reported the identity as 6% in its structure-based sequence alignment. Sequence alignment by MAFFT<ref name="mafft">PMID: 23329690</ref> obtained 18% sequence identity using more and larger gaps. <scene name='87/875686/Dali_5a2f_vs_7jx6_yale/2'>The structural similarity between Dali's top hit and 7jx6</scene><ref name="yale">Structural superposition by Dali. Interpolation by the [http://www2.molmovdb.org/wiki/info/index.php/Morph2_Server Yale Morph2 Server]. Homogenization method: homology modeling. No minimization. This produced a 9-model file where model 1 was 7jx6, and models 2-9 were interpolations. 5a2f residues 28-133 were added as model 10 (black in the molecular scene).</ref> is not as close as for AlphaFold2's prediction, but is closer than the 2nd best prediction (see Table I below). In conclusion, '''ORF8 does not have a novel fold'''<ref name="holm">The interpretation of Dali's result to mean that ORF8 does not have a novel fold was kindly confirmed by Liisa Holm, personal communication to [[User:Eric Martz|Eric Martz]], February, 2021.</ref>.
 ===AlphaFold2 Prediction for ORF8===
@@ Line 33: / Line 33: @@
 | AlphaFold2 || 87 || 3 || 2.58<br>'''1.25''' || 92/92 (100%)<br>'''83/92* (90%)''' || 3.23<br>'''1.91''' || 747/748 (100%)<br>'''679/748 (91%)'''
 |-
-| Dali top hit [[5a2f]] || 53<ref name="gdt_ts" /> || na || 3.2<br>'''1.95''' || 92/92 (100%)<br>'''48/92 (52%)''' || na || na
+| Dali top hit<ref name="nnf">See [[#ORF8 is not a novel fold]].</ref> [[5a2f]] || 53<ref name="gdt_ts" /> || na || 3.2<br>'''1.95''' || 92/92 (100%)<br>'''48/92 (52%)''' || na || na
 |-
 | 2nd Best* || 43 || 0 || 5.33<br>'''1.71''' || 92/92 (100%)<br>'''38/92 (41%)''' || 6.54<br>'''5.86''' || 747/748 (100%)<br>'''324/748 (43%)'''