Click here for legend (bottom of page) or here for single page

1A19 - barstar

27-30: Models adopt alternate loop at crystal contact. Crystal homologs match native and NMR homologs adopt yet another alternate. (V, C, N)

60-80: Models shift dimer-contact helix and adopt alternate loops on both ends. Crystal and some NMR homologs match native; other NMR homologs adopt yet another alternate. Models' helix N-terminus is floppy but NMR homologs' helix N-terminus is floppy in a different way. (V, M, N)

isolated monomer simulation

crystal lattice simulation

1A32 - ribosomal protein S15 (+sc)

22-86: Rosetta's global alternate is not matched by a low-resolution structure of the same protein in the context of more ribosome components, but that structure is low-resolution, Calpha-only, and unreliable, so the alternate remains plausible. (V, C/L, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

25 & 76: Models adopt different Gln & Arg rotamers, most likely because explicit water is missing. (V, C/W, R)

34: MolProbity suggests 180 degree Gln flip to alleviate clash, but models move toward crystal contact instead. No homologs are available to confirm. (V, C/E, N)

36: Models all agree with MolProbity's 180 degree Asn flip. (V, E, S)

39: MolProbity suggests 180 degree Asn flip to alleviate clash, but models move toward crystal contact instead. No homologs are available to confirm. (V, C/E, N)

53: Arg rotamer outlier at surface. Models adopt reasonable alternative. (V, E, N)

69: Models choose alternative to clashing Leu rotamer outlier, but no homologs are available to confirm. (V, E, N)

1A68 - tetramerization domain of shaker potassium channel

134-151: Models shift helix sideways toward crystal contact and away from tetramer contact, likely because adjacent C-terminus is floppy. All iso/homologs match native, but are also tetramers or have extended C-terminus. (V, C/M, N)

isolated monomer simulation

crystal lattice simulation

1A8O - HIV capsid CTD (+sc)

151-220: Two different conformations are apparent for length of full protein. Crystal and dimer contacts could be involved. Not confirmed by alternative X-ray structures, which differ from each other but in different ways. The two globally distinct backbone conformations correspond to two different disulfide "rotamers," neither of which matches the native one. (V, C/M, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

169: Deposited Tyr has crystal contact. No homologs match models' rotamer cluster that differs in just chi2, but some homologs with different crystal packing match models' rotamer cluster that differs in chi1 and chi2 (although with backbone shift). (F, C, S)

175: Models and some homologs agree on alternative Glu rotamer at dimer interface. However, backbone is not 100% converged and other homologs adopt other rotamers. (V, M, S)

188: Models and some homologs agree on alternative Thr rotamer at dimer interface. (V, M, S)

198--218: Models' disulfide rotamer is different from that in deposited structure and all homologs. Probably causes two globally distinct alternate backbones. (F, ?, R)

202: Models flip Leu rotamer, but deposited is not rotamer outlier and all homologs agree with it. (F, ?, R)

211: Many models flip Leu rotamer, but deposited is not rotamer outlier and all homologs agree with it. (F, ?, R)

1ACF - A. castellanii profilin IB (+sc)

1-10: Models are floppy around native at N-terminus with crystal contacts. Seems reasonable, but NMR homolog is floppy in a different way. (F, C, N)

51-57: Models are distinct yet floppy at a helix with crystal contacts. NMR homolog is also floppy but in a different way. X-ray homologs match native closely. (F, C, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

41: MolProbity strongly recommends 180 degree Gln flip. All models with same rotamer agree, despite backbone not fully converging. (F, E, S)

60: Phe rotamer outlier (but just barely). Models and homologs differ from deposited only slightly. (F, E, S)

62: Leu rotamer outlier with clash. Models plausibly flip, and some NMR homolog models match. (F, E, S)

78: Mild deposited Tyr clash. Most models adopt chi1 rotamer totally different from all homologs, but some models adopt different chi2 rotamer matched by some homologs. (F, E, S)

84 & 102: Deposited Ser & Glu rotamers participate in H-bonds. Models adopt different rotamers. Homologs happen to be point mutants to Ala so can't confirm. (F, ?, N)

116: Leu rotamer outlier. Models and some homologs flip. (F, E, S)

1AIL - apo RNA-binding nonstructural (NS1) protein

1-25: Rosetta has alternate(s) for N-terminal helix which forms majority of dimer contact. Dimeric NMR isolog and 82-89% X-ray homologs, including one bound to RNA, all match native. Can't confirm Rosetta as plausible pre-dimerization state. (F, M, N)

30-50: Rosetta has alternate(s) for helix involved in dimer contact. Does not represent RNA-bound state because RNA-bound 89% homolog 2ZKO matches native almost perfectly. May reflect plausible pre-dimerization state but cannot confirm. (F, L/M, N)

isolated monomer simulation

crystal lattice simulation

1AIU - human thioredoxin, D60N mutant, reduced form

18-21 & 80-84: Loops collapse toward each other & away from crystal contact. Supported by NMR but not X-ray homologs. (F, C, S)

29-38: Models are disordered at this dimer interface. Several X-ray and NMR homologs with 90+% identity support the existence of flexibility. Disulfide from beginning to end of loop is formed in neither native nor Rosetta models but is formed in some X-ray homologs and ~all NMR homologs, so it may be transiently formed and somehow relevant. (F, M, S)

isolated monomer simulation

crystal lattice simulation

1B3A - anti-HIV protein AOP-RANTES

14-19: Models are floppy because N-terminal, strand-swapped residues (including disulfides) were not simulated. (F, M, R)

29-38: Loop is flexible in models because an adjacent region of structure was not simulated and thus disulfide 10-34 cannot form. (F, M, R)

isolated monomer simulation

crystal lattice simulation

1BGF - conserved N-terminal domain of STAT-4

9-20 & 75-80: Extensive dimer contact to helix and loops, the lack of which seems to impact Rosetta. No close homologs available to validate. (V, M, N)

45-55: Crystal contact at helix-turn-helix where models deviate. No close homologs available to validate. (V, C, N)

87-97: Crystal contact at helix-loop transition where models deviate. No close homologs available to validate. (V, C, N)

isolated monomer simulation

crystal lattice simulation

1BK2 - alpha-spectrin SH3 domain, D48G mutant

17-22: Models and both types of iso/homologs move away from native Pro-Phe van der Waals crystal contact. Native structure is also suspect due to rotamer outliers, but density is not available to investigate. (F, C, S)

36-40: Models deviate at loop with native crystal contact. NMR homologs deviate similarly. (F, C, S)

45-49: Models move loop toward adjacent native crystal lattice. Large ensemble of X-ray iso/homologs matches quite well. Native structure is also suspect due to Ramachandran and rotamer outliers, but density is not available to investigate. (F, C, S)

isolated monomer simulation

crystal lattice simulation

1BKR - calponin homology domain (+sc) (see also Fig. 8)

54-58: Rosetta moves loop ~8Å into space occupied by native crystal contact, but X-ray isolog and 87% NMR homolog do not match. Simulation in crystal lattice reverts loop to native conformation, so Rosetta's deviation remains plausibly due to crystal environment. (V, C, N)

88-90: Pronounced, ~5Å deviation at loop in crystal contact, but X-ray isolog and 87% NMR homolog do not match. Simulation in crystal lattice does not revert loop to native conformation, so Rosetta's deviation is likely an error. (V, C, R)

isolated monomer simulation

crystal lattice simulation

Sidechains:

77: Most Rosetta models flip Thr rotamer outlier at crystal contact. X-ray isolog disagrees, but has same crystal contact and is also a rotamer outlier so not helpful. Re-refined structure and 87% NMR homolog agree with flip. Thus the crystal contact does not dictate the flipped rotamer; however, its absence does make the sidechain mobile in Rosetta's models. Flipped rotamer makes H-bond to explicit water (B<30); Rosetta prefers it despite using implicit solvent. (F/V, C/E/W, S)

101: Many Rosetta models flip partially solvent-exposed Thr rotamer outlier. Re-refined structure and 87% NMR homolog agree with flip. Flipped rotamer makes H-bond to explicit water (B~10); Rosetta prefers it despite using implicit solvent. (F, E/W, S)

1BM8 - MCB-binding factor (MBF) (+sc)

10-14: Rosetta's beta hairpin is floppy at a crystal contact, but there is no true confirmation. (V, C, N)

19-25: Rosetta tucks beta hairpin more into the protein. Seems reasonable that this area would be flexible, but there is no true confirmation. Ordered waters, also in X-ray isolog, might block floppiness in vivo though. (F, W?, N)

46-48: Rosetta deviates at helix-turn-helix, presumably because crystal contact is not modeled. NMR isolog also deviates but in a different way. (V, C, N)

58-63: Rosetta deviates at C-terminus of helix. Maybe because of crystal contact but maybe in response to nearby Ile82 rotamer change that alleviates native clash (error). NMR isolog also deviates but in a different way. (V, C/E, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

65: Models adopt surface Lys rotamer different from deposited clashing rotamer outlier. X-ray isolog 1MB1 matches models despite having deposited crystal contact, so deviation is due to error, not crystal contact. (V, E, S)

70: Deposited Phe H-bond to water in crystal contact. Models adopt different rotamer. X-ray isolog 1MB1 has quite different backbone despite having deposited crystal contact, so alternates are plausible here. (V, C/W, N)

72: Models adopt Lys rotamer different from deposited rotamer outlier. X-ray isolog 1MB1 matches models despite having deposited crystal contact, so deviation is due to error, not crystal contact. (F, E, S)

82: Models adopt different Ile chi2 from clashing deposited rotamer, but X-ray isolog 1MB1 matches deposited. (V, E, N)

1BMG - bovine beta 2-microglobulin (+sc)

No significant deviations; good match to deposited structure.

isolated monomer simulation

crystal lattice simulation

Sidechains:

8: Models adopt alternative Gln rotamer at water-mediated crystal contact, but no >76% sequence identity homologs exist to confirm. (V, C/W, N)

21: Models adopt plausible alternative to badly clashing Asn rotamer outlier, but no >76% sequence identity homologs exist to confirm. (V, E, N)

24: Deposited error. Models flip Asn 180 degrees, and MolProbity agrees. (V, E, S)

59: Models adopt alternative Trp rotamer at strong crystal contact, but no >76% sequence identity homologs exist to confirm. (V, C, N)

1BQ9 - rubredoxin, formyl methionine mutant (+sc)

2-52: Two very distinct clusters of overall global conformation. Inconclusive since Rosetta doesn't simulate the Fe atom, but suggestive that this protein may fold even in the absence of Fe. (V, ?, R)

isolated monomer simulation

crystal lattice simulation

Sidechains:

9: Deposited Fe-binding Cys has well-defined rotamer, but models adopt variety of rotamers because Fe not modeled. Homologs either have shifted backbone or match deposited. (F, L, N)

25: Crystal contact. About half of models and about half of NMR homologs adopt alternate Ser rotamer. (F, C, S)

29: Models adopt alternative Lys rotamer at crystal contact. X-ray homologs match deposited because also have crystal contact. NMR homolog is floppy and does not quite match deposited or models. (V, C, N)

1C8C - small chromatin protein Sso7d

6-13: NMR structure of K13L mutant doesn't confirm movement of hairpin (which contacts DNA), but then the mutation is very close so it's not fully conclusive. Rosetta deviation seems very reasonable. (F, L, N)

isolated monomer simulation

crystal lattice simulation

1C9O - cold shock protein Bc-Csp (+sc)

19-24: Homologs generally agree with Rosetta's two very reasonable alternates for this loop with a crystal contact and bound sodium. (F, C, S)

52-59: Rosetta finds two alternates, both of which move this hairpin loop with dimer and crystal contacts in the same direction. Many homologs, including apo NMR structures, also deviate but move in the opposite direction. Rosetta's monomeric form is unlikely. (F, C/M, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Sidechains:

8: Models mostly adopt alternative to Trp rotamer at crystal contact. Homologs with similar lattice to deposited match deposited. (F, C, N)

11: Models and MolProbity flip Asn 180 degrees to alleviate clash, but models also alter chi1 to move into deposited crystal contact; no homologs quite match. (F, C/E, N)

50: Models adopt various alternatives to mildly clashing Glu rotamer at crystal contact. Homolog 1I5F with very slightly different lattice matches one of them nicely. (F, C/E/W, S)

53: Models have alternate rotamers as well as backbone shifts at dimer interface / crystal contact. No homologs validate models because they too are dimers and have crystal contacts. (F, C/M, N)

1CC8 - Atx1 metallochaperone (+sc)

55-73: Discrete, crisp alternate backbone for final 1/3 of sequence. Strange, but well within the NMR ensemble. Some crystal contacts may play a role, but Hg & benzamide ligands are distal and unrelated. (V, C, N)

(One of the X-ray isologs (1CC7) has an alternative conformation in loop 13-18, at the Hg/benzamide binding site, but Rosetta does not locate this minimum within 3 energy units of its minimum-energy model.)

isolated monomer simulation

crystal lattice simulation

Sidechains:

6: Models all flip His 180 degrees. Classic MolProbity example! (V, E, S)

8: Of models with deposited rotamer, more flip Gln 180 degrees than do not. MolProbity mildly suggests flip on basis of explicit waters, so Rosetta gets it somewhat right based on implicit solvent but could maybe do better with explicit solvent. (F, E?/W, N)

30 & 36: Benzamidine from crystallization medium dictates Glu & Ile rotamers for deposited and X-ray homolog. Models and some apo NMR homolog models agree on alternative. (V, L, N)

46: Asp rotamer clashes with His6 and models reasonably rotate chi2, but no homologs exactly match models. (V, E, N)

71: Models and NMR homolog adopt range of rotamers instead of clashing Lys rotamer outlier at crystal contact to mercury-bound site of another protein in the lattice. (F, C/E, S)

73: Models, X-ray homolog from different lattice, and NMR homolog mostly agree on alternative to clashing deposited Leu rotamer. (F, C/E, S)

54: Models agree on alternate surface Asp rotamer for no obvious reason. NMR homologs do commonly adopt a rotamer with the models' chi1 but not with the models' chi2. (V, ?, R)

58: Surface Glu rotamer clashes. Models tweak chi2 or switch to a totally different rotamer, but no homologs exactly match models. (F, E?, N)

1CEI - colicin E7 immunity protein ImmE7

3-87: Lowest energy conformations are essentially globally correct, but very nearby are conformations which are significantly different. This might be due to the missing dimer -- maybe this protein doesn't comfortably fold until it pairs up with a partner, although there is no experimental evidence for this. (F, M, N)

50-56: On opposite end of protein from dimer interface, Rosetta shifts helix toward native crystal contact. X-ray homologs do not support the deviation but also have crystal contacts (though different from native). (V, C, N)

isolated monomer simulation

crystal lattice simulation

1CG5 - D. akajei (stingray) hemoglobin, deoxy form

Rosetta models are globally unsettled in the absence of hemes and tetramer partners.

40-49 & 79-93: Especially extensive deviations at heme binding site. These alternate states are reasonable but cannot be confirmed in detail -- e.g. NMR relaxation data would be needed to "confirm" floppiness. (F, L/M, N)

114-120: More pronounced deviation because crystal contact is not modeled in addition to lack of heme and tetramer partners elsewhere. (F, C, N)

isolated monomer simulation

crystal lattice simulation

1CTF - ribosomal protein L7/L12

62-64: Rosetta shifts SO4-binding loop like a hinge. NMR and X-ray iso/homologs almost unanimously agree. (V, L, S)

75-89: Helix in dimer interface; SO4 binding site at C-terminus. NMR isologs disagree with Rosetta's deviation. (V, M/L, R)

97-109: Helix with crystal contacts distal from dimer interface and SO4 binding site. NMR isologs disagree with native to about the same extent that Rosetta's models disagree with native, but in a different way. (V, C, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1DHN - dihydroneopterin aldolase (see also Fig. 3a)

39-50: Cannot confirm flexibility for loop (also influenced by crystal contacts) in putative monomer, but deviation is significantly reduced when simulated in crystal. (F, C/M, N)

100-111: Cannot confirm loop flexibility in putative monomer, but deviation is significantly reduced when simulated in crystal. (F, M, N)

isolated monomer simulation

crystal lattice simulation

1DZO - PAK pilin

Most of the structure is fairly well converged to near native, with some 3-3Å deviants within 3 energy units of the minimum energy model.

62-72: Rosetta moves this somewhat strange hairpin loop like a hinge. Crystal contacts seem involved, but several X-ray iso/homologs have different lattices and vary less, so Rosetta's flexibility is plausible but may be excessive. (F, C, N)

133-143: Rosetta models this disulfide-flanked loop as flexible. The 98% identical X-ray homolog 2QY0 also deviates here and has different crystal lattice from native, but also has complicating local T130K and E135K mutations. Intriguingly, this loop is the receptor-binding site in the polymerized form of this protein (bacterial pilin) and thus has a functional role of medical interest! (F, M, N)

isolated monomer simulation

crystal lattice simulation

1E6I - Gcn5p bromodomain + acetylated H4 peptide

354-371: Loop lining binding groove is extremely flexible, as is the entire protein to some extent. All seems very reasonable, but no homologs or any other supporting or disproving data. (F, L, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1ELW - Hop TPR1 domain + HSC70 peptide

10-18 & 101-118: Very minor collapse of two most exposed helices onto peptide binding groove. Seems very reasonable, but no homologs exist to confirm. (F, L, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1ENH - Drosophila engrailed homeodomain (see also Fig. S2a)

1-7: Models floppy but distinct from native DNA-free crystal structure at N-terminus. Alternate is not DNA-bound state because DNA-bound X-ray isolog 1HDD deviates in a different way at N-terminus. Both Rosetta's models and the NMR isolog 2JWT deviate significantly, so crystal contacts may influence 1ENH; however, Rosetta's helical alternate is quite different from 2JWT's "swath" and therefore cannot be confirmed in detail. (F, C?/L?, N)

isolated monomer simulation

crystal lattice simulation

1EW4 - E. coli frataxin-like CyaY (see also Fig. S2b)

25-28: Some models add a turn to helix preceding this loop. X-ray isologs match native, probably because of same crystal contact, but NMR isolog also much closer to native. (V, C, N)

71-78: Models slightly disordered in loop. NMR isolog is mobile and X-ray isologs disagree with each other, probably because one can (2P1X) can bind Eu metal here! (F, C/L, S)

isolated monomer simulation

crystal lattice simulation

1EYV - transcriptional antitermination protein NusB

37-50: Apparently dimer-induced deviation, but cannot validate because no homologs or electron density are available. (F, M, N)

78-82: Rosetta hinges this loop, which has high B-factors in the native crystal structure. The only crystal contact is weak and water-mediated. Motion for this loop in vivo is very plausible, but cannot be confirmed because no homologs or electron density are available. (F, ?, N)

isolated monomer simulation

crystal lattice simulation

1FAA - spinach thioredoxin (see also Fig. 5)

1-16: N-terminus clearly held in place by beta-sheet-like crystal contact, which explains Rosetta's disorder and shift toward the body of the protein. All three iso/homologs are truncated to at least residue 10, probably because floppiness prevented crystallization. Furthermore, in-lattice simulation stabilizes this region. (F, C, S)

33 & 96-102: Rosetta shifts Gly33 and this loop toward each other. The loop has a significant crystal contact, but the iso/homologs (2/3 of which have different space groups from 1FAA) match the deposited structure. Looks suspiciously like a Rosetta error. (V, C, R)

isolated monomer simulation

crystal lattice simulation

1FKB - human immunophilin FKBP12 + rapamycin (see also Fig. 6)

31-33: Crystal contact appears to bias native loop. Many crystal isolog structures match Rosetta's alternate. (V, C, S)

40-45: Rosetta chooses a distinct yet floppy alternate for this loop with crystal contacts near rapamycin binding site. Essentially all X-ray isologs match native. NMR are floppy, but don't quite superimpose onto Rosetta's models. (F, C/L, N)

56-62: Rosetta "shears" helix in contact with rapamycin ligand and in crystal contact. X-ray isologs match native, but NMR isologs are somewhere in between, perhaps closer to Rosetta's models. (V, C/L, S)

isolated monomer simulation

crystal lattice simulation

1FNA - cell adhesion module of fibronectin

9-14: Half of Rosetta's models match the deposited structure, but the other half split out into a discrete alternate. Of the four domains of isolog 1FNF, a single chain with four copies of the 1FNA domain, one matches Rosetta's alternate almost perfectly, a second matches the deposited structure almost perfectly, and the other two are halfway in between. There is a strong crystal contact in 1FNA, but 1FNF has a different lattice. Furthermore, 1FNA Val11 clashes with Leu19, a rotamer outlier on the adjacent strand; Rosetta appears to fix this error and is corroborated by several iso/homologs. Nice confirmation of the computed alternate! (V, C/E, S)

40-44: Definite crystal contact. Iso/homologs confirm the existence of Rosetta's less pronounced loop flexibility, but not really the exact conformations it chooses. (F, C, N)

75-83: Loop has strong crystal contact. Severe disorder exhibited by Rosetta models and NMR iso/homologs; some disorder in X-ray iso/homologs. Rosetta also fixes a pre-Pro Ramachandran outlier for Ser81-Pro82. (F, C/E, S)

isolated monomer simulation

crystal lattice simulation

1GVP - ssPA-binding gene V protein (see also Fig. 3b)

35-41: Rosetta deviates at loop with both dimer and crystal contacts. The only X-ray homolog with a non-native space group (1YHA) and both NMR homologs, all of which are dimeric, match Rosetta well, suggesting the crystal contact is a bigger influence here than the dimer contact. (F, C/M, S)

49-54: Loop deviation at crystal contact. X-ray structures match native due to similar crystal contacts, but Rosetta's alternate is confirmed by NMR homologs 2GVA and 2GVB. (F/V, C, S)

62-82: Huge 15Å loop swing at dimer interface. Homologs are all dimers too so can't be used to confirm this as a plausible pre-dimerization monomeric state. (F/V, M, N)

isolated monomer simulation

crystal lattice simulation

1H75 - glutaredoxin-like Nrdh-redoxin

8-11: Beta-turn-helix clearly pulled by crystal contact, but no homologs available to confirm. (V, C, N)

69-76: C-terminus disordered, apparently because crystal contact is not modeled, but no homologs available to confirm. (F, C, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1HZ6 - B1 domain of P. magnus protein L

4-64: Most low-energy models are very close to native, but a significant subset is non-native-like. X-ray isolog and >90% sequence identity X-ray and NMR homologs match native closely, so Rosetta's alternate is likely an error. Not modeling the six N-terminal residues seems not to be a factor. (F, ?, R)

13-16: For globally native-like subset, Rosetta swings out loop with crystal contact. Deviation is matched precisely by NMR homolog, and somewhat by X-ray isolog and homologs with non-native crystal contacts. (V, C, S)

19-23: For globally native-like subset, Rosetta subtly shifts beta sheet in very strong sheet-like crystal contact. NMR homolog matches well. (F, C, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1IG5 - intracellular Ca(2+)-receptor calbindin D9k

1-75: Entire structure is somewhat non-convergent, probably due to the missing Mg atom. (F, ?, R)

39-45: Beyond global non-convergence, this loop is especially displaced from the deposited structure. Most iso/homologs, even several metal-bound, match the deviation. Obviously crystal-contact-induced. (V, C, S)

isolated monomer simulation

crystal lattice simulation

1IIB - E. coli IIBcellobiose

42-49: Helix shifts slightly perpendicular to crystal contact. Leu70 switches rotamer to make room, which some NMR iso/homologs match. However, no iso/homologs exactly match Rosetta's backbone despite disagreement amongst themselves. (V, C, N)

95-105: Tail shifts slightly away from crystal contact. NMR iso/homologs are flexible but mostly move in the other direction. (F, C, N)

isolated monomer simulation

crystal lattice simulation

1JFV - S. aureus arsenate reductase, oxidized C10S/C15A mutant (see also Fig. 10)

82-98: Fascinating alternate with biological connection! The deposited structure binds perchlorate nearby and forms SS 82-89, which is the oxidized end point of a Cys10-Cys82-Cys89 reaction cycle (here mimicked by a C10S mutation). A minority of Rosetta models form SS 82-89 and thus match the native loop, which has some rotamer/Ramachandran/clash errors. However, a sulfate-bound isolog and the majority of Rosetta models adopt a mini-helix instead. Though the balance is dependent on Rosetta's SS strength parameter, the current results suggest that this SS bond is only transiently formed -- in accord with its demonstrated enzymatic function! (Point mutants of various cysteines, including some bound to perchlorate, further confirm this alternate. Also, a minor crystal contact to the loop is probably not a major factor.) (V, L/C/E, S)

isolated monomer simulation

crystal lattice simulation

1KPE - protein kinase C interacting protein

14-29: A small minority of low-energy models are variable due to a crystal contact, but that cannot be validated. (F, C, N)

102-105: A small minority of low-energy models are variable at this dimer contact, but that cannot be validated. (F, M, N)

106-108: Rosetta models' backrub-like swing is pretty well matched by all isologs, including several bound to adenosine variant ligands, despite not simulating a ligand. (V, L, S)

isolated monomer simulation

crystal lattice simulation

1L6P - N-terminus of E. coli DsbD (see also Fig. S2c)

4-10: N-terminus is floppy in models and in X-ray isologs, 3/4 of which have non-native crystal contacts. Tail is clearly flexible. (V, C, S)

69-72: Models pull loop toward main structure at native crystal contact. Non-native crystal contacts push the loop the other direction in 3/4 X-ray isologs. So loop is probably flexible in vivo. (V, C, N)

isolated monomer simulation

crystal lattice simulation

1LOU - ribosomal protein S6 (+sc)

The deposited structure is an isolated version of a ribosomal protein. The fact that Rosetta models computed either in isolation or in a simulated lattice, as well as distal point mutants, match extremely well suggests that ribosome assembly doesn't affect this protein's structure very much. This is likely because it is located on the surface of the ribosome rather than penetrating into its deeper recesses like many other ribosome-associated proteins.

67-74: Rosetta slightly compresses the N-terminus of this 310 helix. The deposited structure has a crystal contact, but several distal point mutants bind RNA as part of the entire ribosomal subunit instead and thus cannot be used to confirm the plausible computed alternate. (F, C/L, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

16: Models adopt plausible alternative to Gln rotamer outlier at crystal contact, but no homologs match. (V, C/E, N)

17: Models and some homologs match models' alternative to Ser rotamer outlier at crystal contact. (V, C/E, S)

18: Some homologs match one of models' alternatives to clashing Gln rotamer. (F, E, S)

27: Models adopt plausible alternative to surface Gln rotamer outlier, but no homologs match. (V, E, N)

28: Deposited Arg contacts intra-chain water as well as crystal contact. Models adopt different set of Arg rotamers. Homolog 1CQN lacks water and crystal contact and basically matches models. (F, C/W, S)

50 & 52: Models adopt alternative Ile & Tyr rotamers at water-mediated crystal contact, but homologs have deposited lattice so match deposited. (V, C/W, N)

77: Deposited surface Arg is rotamer outlier. Models are floppy; homologs are too, but in a different way. (F, E, N)

1MGW - RNase Sa3, cytotoxic microbial ribonuclease (+sc)

1-6: Rosetta adopts alternate for N-terminus -- flexible but very distinct from native -- and isolog 1MGR (with different crystal contacts) agrees. (F, C, S)

62-69: Loop adopts two flexible alternates. Some X-ray homologs and especially the 71% identical NMR homolog 1C54 agree there is variability but have local sequence changes. Reasonable but cannot confirm. (F, ?, N)

78-81: Rosetta collapses lithium-binding loop into pocket, and apo isolog 1MGR and 72% identical homolog 1T2I agree. (V, L, S)

isolated monomer simulation

crystal lattice simulation

Sidechains:

73: Models adopt alternative core Ile rotamer for no obvious reason. Isolog 1MGR matches deposited. Homolog 1T2I matches model rotamer although not backbone. (V, ?, S)

75: Deposited H-bond to two explicit waters. Models unanimously choose alternative rotamer, but isolog 1MGR also has waters and matches deposited. (V, W, R)

94: Models adopt two alternative Arg rotamers. Isolog 1MGR matches one of them; other would conflict with explicit water (B<40) in deposited and 1MGR. (F, E/W, S/R)

1MJC - major E. coli cold shock protein CspA

2-6: The deposited N-terminus is held in place by a crystal contact, but the NMR isolog does not match Rosetta's deviation. Flexibility probably exists here but Rosetta is not modeling it properly. (F, C, R)

23-29: Subset of ensemble from NMR isolog matches Rosetta's flexible loop at a crystal contact. (F, C, S)

35-44: Rosetta predicts two alternates in the absence of the crystal contact from the deposited structure. The NMR isolog confirms flexibility somewhat but does not exactly coincide with the less helical alternate. There is no support for the more helical alternate. MolProbity errors (rotamers, Ramachandran, sterics) and weak electron density cast some doubt on the deposited loop conformation. (F/V, C/E, N/R)

56-63: NMR isolog basically encompasses Rosetta's flexible loop at a crystal contact. Probably close enough to be considered "supported." (F, C, S)

isolated monomer simulation

crystal lattice simulation

1NPS - N-terminal domain of Ca-binding protein S

Rosetta's models are nearly a perfect match for essentially the entire protein, indicating that calcium binding causes little to no conformational change.

25-33: Computed low-energy models shift this surface helix away into a crystal contact that is not simulated. The NMR isologs match the alternate impressively closely. (F, C, S)

isolated monomer simulation

crystal lattice simulation

1NPU - murine PD-1 (+sc)

37-42: Rosetta's flexibility for this loop at a crystal contact is validated by iso/homologs with various contacts. (F, C, S)

57-59: Rosetta shifts loop toward space occupied by ordered water (B=16) with 3 H-bonds in native structure. The native Gln58 sidechain is also involved in a crystal contact, but both isologs (with various contacts) match native anyway. Rosetta may be wrong. (V, C/W, R)

66-71: Rosetta shifts the loop away from a crystal contact, but all X-ray iso/homologs (with various contacts) match native. This deviation may be too extensive but still seems pretty reasonable. (F, C, N)

95-99: Native loop has weak electron density, high B-factors, and MolProbity Ramachandran outliers. Both isologs (with various crystal contacts) match Rosetta's extent of floppy deviation. (F, C/E, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Sidechains:

8 & 10: Models and homolog 3BIK adopt alternative Trp & Thr rotamers at crystal contact. (F, C, S)

16: Models and homologs adopt alternative to clashing Asn rotamer outlier at crystal contact. (V, C, S)

23: Models and homologs flip clashing Leu rotamer outlier. (F, E, S)

31: Models adopt floppy set of Met rotamers different from deposited rotamer outlier at crystal contact. Homologs are also floppy but don't exactly match models. (F, C/E, N)

33: Deposited Asn rotamer influenced by adjacent sidechain altered by crystal contact. MolProbity flips 180 degrees to alleviate clash, but models change chi2 instead. Homologs are different still because of another more direct crystal contact. (F, C/E, N)

63: Minor model population and some homologs adopt alternative to clashing Gln rotamer at crystal contact. Ordered water from deposited-specific crystal lattice also involved. (F, C/E/W, S)

75: Models adopt Met rotamer different from deposited with clashes to adjacent disulfide, but homologs match deposited. (F, E?, N)

1O4W - A. fulgidus PilT N-terminus domain

122-134: Extremely floppy tail. Very reasonable since tail forms >10-residue-long strand in native dimer, but cannot confirm. (F, M, N)

isolated monomer simulation

crystal lattice simulation

1O5U - T. maritima enzyme TM1112 (see also Fig. S2d)

8-20: Rosetta chooses different position for helix in crystal contact. NMR isolog 2K9Z matches models, but NMR isolog 1LKN matches native. (V, C, S)

35-38: Rosetta mildly shifts beta-turn near unknown ligand binding site. Apo NMR homologs, especially 2K9Z, are closer to models than to native. Probably due to crystal contact but ligand interaction is not out of the question. (V, C/L?, S)

isolated monomer simulation

crystal lattice simulation

1O73 - T. brucei tryparedoxin

15-21: Plausible, minor backbone deviation for loop with some clashes, a MolProbity rotamer outlier, and a crystal contact, but no confirmation. (F, C/E, N)

126-139: Plausible, minor backbone deviation for loop with crystal contact, but no confirmation. (V, C, N)

isolated monomer simulation

crystal lattice simulation

1OPD - histidine-containing protein HPr, S46D phosphorylation mimic mutation (+sc)

9-12: Rosetta swings out loop at crystal contact. X-ray homolog 3CCD with non-native crystal lattice matches fairly well. (F, C, S)

46-54: Rosetta adopts a slightly squeezed helix conformation which cannot be validated by >60% homologs. (F, C/?, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Sidechains:

4: Mild Gln clashes. MolProbity mildly recommends 180 degree flip (but doesn't know about crystal contact). Models adopt three alternate rotamers instead; all are matched by at least one NMR homolog or crystal homolog with different lattice lacking local crystal contact. (V, C/E, S)

14: Models and homologs flip Leu rotamer outlier. (V, E, S)

30: Models and homologs flip Thr rotamer outlier. (V, C/E, S)

57: Models and MolProbity flip Gln 180 degrees because neither sees crystal contact including extra water. Crystal homologs with same contact match deposited but some NMR homologs match models. (V, C/W/E, S)

74: Models and homologs flip mildly clashing core Val sidechain in bad part of Fo-Fc difference density in order to better fit density. (V, E, S)

77: Models and homologs flip core Leu rotamer outlier in bad part of Fo-Fc difference density in order to better fit density. (V, E, S)

1PGX - B2 immunoglobulin-binding domain of streptococcal protein G (+sc)

Two distinct global conformations separated by approximately a 1/2 register shift. None of the best sequence matches really confirm.

54-69: Most deviant region is the final two beta strands. Trp56 gets exposed in the more deviant cluster (Calpha RMSD ~2A). No X-ray homologs match. NMR isologs differ from native at 58-62 turn, but in opposite direction from models. Interesting but cannot confirm. (V, C, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

21: Crystal contact via explicit water. Models and many homologs adopt alternative Asn rotamer. (V, C/W, S)

25: Deposited alternate conformations A and B. Models and homologs mostly choose alt B. (F, ?, ?)

53: Models and many homologs adopt alternative Asp rotamer at crystal contact. Waters are involved but homologs confirm deviation is not due to implicit solvent failing. (V, C/W, S)

68: Models and many NMR homologs adopt alternative Thr rotamer at crystal contact. Waters involved but homologs confirm deviation is not due to implicit solvent failing. (V, C/W, S)

1POH - histidine-containing protein HPr, wildtype

15-30: Helix slides around slightly, and interestingly His15 can be phosphorylated as part of a signaling pathway, but no homologs match Rosetta. Helix could indeed be flexible to some extent, but not to the degree suggested by Rosetta. Most likely an energy function problem. (F, C/?, R)

46-54: Low-energy Rosetta models all adopt a slightly squeezed helix with altered phi,psi at Gly54 C-cap. The alternate's local Gly54 C-cap conformation is matched by the Ser46Asp point mutant 1OPD, also a target for this study. Rosetta probably chooses this local conformation because its Gly54 phi,psi angles are more common than the native structure's. However, the overall helix position cannot be validated by >60% homologs. (F, C/?, R)

isolated monomer simulation

crystal lattice simulation

1PRQ - profilin-I (+sc)

25-29 & 39-50: Mild coupled deviations are not confirmed by X-ray homologs. NMR homolog shows large amount of flexibility, but that could be due to lack of experimental restraints. (V, ?, N)

50-57: Mild shift of helix at crystal contact is not confirmed by X-ray homologs. NMR homolog shows large amount of flexibility, but that could be due to lack of experimental restraints. (V, C, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

3: Models almost unanimously adopt alternative Gln rotamer at crystal contact, but X-ray homologs match deposited (because same lattice?) and NMR homolog is too floppy to help either way. (F, C, N)

41: Models and MolProbity flip clashing Gln 180 degrees. (V, E, S)

55: Models and 2/3 X-ray homologs adopt alternative to clashing Ile rotamer. (F, E, S)

80: Models adopt one of two alternatives to clashing Lys rotamer, but no homologs quite match. (F, E, N)

96: Models almost unanimously adopt alternative to deposited (very) mildly clashing Leu rotamer. X-ray homologs match deposited and NMR homolog is too floppy. No density is available to confirm classic 180 degree flip. (V, E, N)

105: Models mostly adopt alternative to deposited Asn rotamer outlier, but no homologs quite match. (V, E, N)

116: Models and 2/3 X-ray homologs adopt alternative to deposited Leu rotamer outlier. Classic 180 degree flip. (V, E, S)

1PTQ - activator-binding domain of protein kinase C delta

252-256: Rosetta's loop hinge at crystal (dimer?) interface is matched by NMR homolog despite local sequence difference. Reasonable deviation at biologically significant binding site. (F, C/M, S)

269-280: Native His269 and Cys280 bind separate Zn atoms, but Rosetta's C-terminus is disordered in the absence of Zn. X-ray isolog and NMR homolog are Zn-bound and match native very well. (F, L, R)

isolated monomer simulation

crystal lattice simulation

1PY9 - myelin-oligodendrocyte glycoprotein extracellular domain

7-12: Rosetta hinges intermediate portion of N-terminal beta strand that sticks out of plane of sheet. Region participates in dimer contacts in native and crystal contacts in deposited structure. X-ray homologs with a few (distal) point mutations match Rosetta. (F, C/M, S)

102-105: Some models swing exposed loop away from crystal contacts and crystallographic sulfate. Reasonable, but homologs do not support. (F, C/L, N)

isolated monomer simulation

crystal lattice simulation

1R26 - T. brucei brucei thioredoxin

-7-2: N-terminal tail tucks in to add strand to central sheet, but is held firmly in place by crystal contact in deposited structure. Reasonable but no confirmation. (V, C, N)

isolated monomer simulation

crystal lattice simulation

1R69 - N-terminal domain of phage 434 repressor (see also Fig. S2e)

16-25: Helix shifted at crystal contact, but NMR isologs pretty close to native. (V, C, N)

36-44: Rosetta's alternate at loop is not DNA-bound state because DNA-bound X-ray isologs match native. Likely due to crystal contact instead, since NMR isologs are floppy. (V, C/L, S)

60-63: C-terminus floppy because crystal contact is not modeled by Rosetta. NMR isologs also floppy. (F, C, S)

isolated monomer simulation

crystal lattice simulation

1RKI - P. aerophilum pag5_736 (+sc)

Rosetta forms all 3 disulfides properly in this novel, SS-rich (but not extracellular), thermophilic protein.

11-16 & 38-40: Minor loop deviations at lattice contact. Homologs and electron density are unavailable for validation, but 1RKI chain B (copy of chain A related by non-crystallographic symmetry) matches well! (F, C/M?, S)

96-101: C-terminus collapses inward, likely because deposited crystal contact is not simulated. Homologs and electron density are unavailable for validation, but 1RKI chain B (copy of chain A related by non-crystallographic symmetry) is disordered beyond residue 97, supporting Rosetta's relative disorder. (F, C, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Sidechains:

6 & 88: Ile-Ile clash. Models pack differently, but no homologs exist to confirm. (F, E, N)

12: Models adopt favorable-in-isolation tttt Lys rotamer at crystal contact, but no homologs exist to confirm. (F, C, N)

22--34: Most models adopt alternative disulfide rotamer, but no homologs exist to confirm. (F, ?, N)

24--54: About half of models adopt alternative disulfide rotamer, but no homologs exist to confirm. (F, ?, N)

49, 53, & 56: Surface sidechains near 24--54 disulfide. Rotamer outliers and/or clashes. Models adopt plausible alternatives, but no homologs exist to confirm. (F, E, N)

66: Many models flip Asn 180 degrees. MolProbity agrees. (F, E, S)

1RNB - barnase + d(GpC)

6-20: Rosetta has very floppy N-terminal helix, but X-ray homologs are very close to native despite variety of crystal contacts. (F, C, R)

66-69: Rosetta is variable at a loop where X-ray homologs disagree with each other, likely because a variety of crystal contacts are made and/or crystallization substrates are bound. MolProbity rotamer outliers and Cbeta deviations cast further doubt on the native loop conformation. (F, C/L, S)

77-81: Subset of models have loop deviation for no obvious reason. dGC-binding site is adjacent but seemingly not involved because even apo X-ray homologs match native loop. (V, L, R)

isolated monomer simulation

crystal lattice simulation

1S12 - hypothetical protein TM1457 (+sc)

46-53: Sheet hairpin appears disordered. Mild crystal contact may also bias deposited structure. Ramachandran outlier casts some small doubt on the deposited conformation, but Rosetta's changes are beyond the scope of misfitting within the density. Chains B, C, and D of 1S12 all match chain A precisely, though this could be due in part to simultaneous refinement. Seems reasonable, but no confirmation. (F, C, N)

89-94: C-terminus appears disordered. High crystal B-factors and steric clashes indicate this tail is difficult to pin down. Chains B, C, and D of 1S12 are disordered here, supporting the validity of Rosetta's conformations. (F, C/E, S)

The two helices match quite closely in the monomer simulation, despite forming the dimer interface in vivo, suggesting this dimer interface is essentially pre-formed (unlike others in this data set).

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Sidechains:

62: Cys does not form cross-dimer disulfide bond because other half is not modeled. (V, M, N)

70: Deposited Met clashes badly. Models adopt alternative rotamer that moves into positive difference density. (F, E, S)

84: Some models flip Gln 180 degrees to alleviate deposited clashes. Supported by Probe and MolProbity. (F, E, S)

88: Models adopt ensemble of alternative Gln rotamers because lattice interactions, including explicit water, do not rigidify any particular one. (F, C/W, N)

1SCJ - propeptide + subtilisin E (+sc)

318-326: Loop entering helix appears to have two very slightly different conformations. Deposited structure has high B-factors, rotamer/Ramachandran outliers, and weak density, suggesting some measure of disorder and thus the possibility that alternates like Rosetta's exist. Crystal contact also present, interestingly to the same loop from an adjacent molecule. No direct confirmation possible since no 100% isologs. 87% homolog 1SPB matches fairly well, but has a few local mutations which confound the issue. Enough overall evidence to "support." (V, E, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Sidechains:

332: Models choose different Gln rotamer (with small backbone shift too) at crystal contact. (V, C, N)

333: Deposited Lys has minor clashes, but major factor is crystal contact. Models adopt alternative rotamer that pokes out toward lattice. Homolog 1SPB matches models. (V, C/E, S)

345: Models flip sidechain end 180 degrees to alleviate clashes and form H-bond with water. Fantastic example of classic MolProbity/Probe Asn 180 degree flip. (V, E/W, S)

350: Deposited Thr sidechain clashes with adjacent beta strand. Models and homolog 1SPB agree on a different rotamer. (V, E, S)

361: Models choose different, fully extended extended (tttt) Lys rotamer at crystal contact. (V, C, N)

369: Models adopt two different Glu rotamers from deposited because dimer is not modeled. Homolog 1SPB matches deposited because it is also a dimer. (F, M, N)

1SHF - Src-homology 3 (SH3) domain in human Fyn (see also Fig. S2f)

96-98: Rosetta's alternate loop at crystal contact is matched by X-ray isologs with different lattices. (V, C, S)

114-120: Two different loop conformations are chosen by Rosetta. X-ray isologs 1G83, 1FYN, and 1M27 (with different lattices) vary here considerably, confirming both conformations do occur. The NMR ensembles, one of which uses this loop to bind an oligopeptide, further confirm the variation. NB: The authors say this contact could be part of a biological dimer. (V, C/M?, S)

135-141: Loop "smear" disproved by X-ray and NMR isologs. (F, C, R)

140-142: C-terminus is disordered in Rosetta models. X-ray isologs indicate possible flexibility. (F, ?, N)

isolated monomer simulation

crystal lattice simulation

1SRR - phosphotransferase Spo0F

55-61: Small subset of low-energy models have markedly different loop conformation, a few in each hinge direction. Probably due to lack of calcium coordinated by Lys56 backbone carbonyl. NMR isologs and X-ray structure 1NAT of distal point mutant, unbound to calcium, match one alternate, but the other is completely unconfirmed. (V, L, S/N)

83-96: Entire helix is rotated about 20 degrees, though somewhat flexible around that distinct equilibrium conformation. Crystal contacts in deposited structure prevent such a change. NMR and X-ray iso/homologs show some disagreement in this region, but do not validate Rosetta's alternate. Distal point mutant 1NAT also has crystal contact, though different from 1SRR. No confirmation, not even from NMR, so most likely a Rosetta error. (F/V, C, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1T2I - RNase Sa

39-43: Loop is flexible in models. Many ordered waters close by. Distal point mutant X-ray structures match deposited structure precisely, but have same space group and likely crystal contacts of their own. Rosetta's flexibility possible but cannot be confirmed. (F, C/W, N)

57-66: Loop is flexible in models. Distal point mutant X-ray structures (some bound to nucleic acid or sulfate and/or with different crystal contacts) also differ from deposited structure, but not nearly as much. Deviation reasonable but probably exaggerated. (F, C, N)

93-96: Rosetta does not form disulfide that ties C-terminal residue to body of protein, and is thus mildly disordered. Homologs all form disulfide and match deposited structure. Modeling error. (F, _, R)

isolated monomer simulation

crystal lattice simulation

1TEN - fibronectin type III domain (see also Fig. 1 inset)

42-44: Rosetta models swing loop away from crystal contact. X-ray homologs have different crystal contacts and disagree amongst themselves on the loop position, but largely move it in the opposite direction from Rosetta. Note: X-ray homologs are structures of redesigns of a different loop, so the sequence changes are distal to this loop. (V, C, N)

isolated monomer simulation

crystal lattice simulation

1THX - thioredoxin-2

1-5: Only significant deviation is at N-terminus. Held in contact by crystal contact in deposited structure; also weak density and some higher B-factors. Rosetta's disorder is plausible, but cannot be confirmed since no homologs are available. (F, C, N)

isolated monomer simulation

crystal lattice simulation

1TIF - translational initiation factor IF3, domain 1

3-7: Only significant deviation is at N-terminus. Interacts with nearby long C-terminal helix via aromatic stacking and salt bridges in deposited structure; both are somewhat held in place by crystal contacts. Rosetta doesn't model this C-terminal helix under the assumption that it would be disordered in solution. However, its first third or so has low B-factors and likely stabilizes the N-terminal tail in question. Best categorized as a modeling error. (F, C, R)

isolated monomer simulation

crystal lattice simulation

1TIG - translational initiation factor IF3, domain 2

88-105: Small subset of low-energy models deviate over stretch of sheet-turn-helix. Seems crystal-contact-induced. Crystallographic water also occludes alternate loop. Plausible but cannot confirm. (V, C/W, N)

121-129: Apparently crystal-contact-induced, large loop hinge. Deviation largely disappears when lattice neighbors are simulated. Crystal contact does not occlude alternate, but rather stabilizes deposited conformation. (V, C, S)

isolated monomer simulation

crystal lattice simulation

1TIT - immunoglobulin-like modules from titin I-band, minimized average NMR

39-47: Significant variability at loop. 1TIT is minimized average structure from NMR ensemble 1TIU, which is also variable here. Deposited structure also has numerous MolProbity rotamer/Ramachandran/steric errors, suggesting substantial degree of disorder. Rosetta plausible but not truly confirmed -- or even confirmable without NMR relaxation data. (F, _, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1TTZ - genomics target X. campestris XcR50 (see also Other Examples)

8-10: Rosetta moves this surface loop away from the pull of an H-bond to the lattice. Confirmed by NMR isolog 1XPV. (V, C, S)

63-76: Most low energy models adopt variable but radically different position for C-terminal helix with strong lattice contacts, but NMR isolog 1XPV matches 1TTZ. Rosetta tries to shove Trp71 into the same space it occupies in 1TTZ, resulting in a MolProbity rotamer outlier. Modeling error! (F, C, R)

isolated monomer simulation

crystal lattice simulation

1TUL - TLP20 from baculovirus A. californica

No significant deviations!

isolated monomer simulation

crystal lattice simulation

1UBI - ubiquitin

Overall very well converged to deposited crystal structure.

31-41: C-terminus of helix and subsequent loop move a bit. Certainly within the ensemble of other crystal and NMR structures. (V, C, S)

isolated monomer simulation

crystal lattice simulation

1UGH - human uracil-PA glycosylase

3-18: Rosetta models the N-terminal tail as extremely floppy, but all X-ray isologs match native. In 1UGH Lys14 forms a hydrogen bond network with Asp48, Glu49, and Ser50 to hold the tail down, so the error may be with Rosetta's ionic bond potential. Another source of error is that Rosetta doesn't explicitly model waters like water 71 (B=41.6) which interrupts beta-pairing of 14-19 and 44-49 to push the tail outward; Rosetta's models instead shove 14-19 in to occupy this space. (F, W, R)

48-52: Low energy models are variable at this loop adjacent to the N-terminal tail and with a lattice contact. The Rosetta ensemble and the ensemble of X-ray structures exhibit similar degrees of flexibility but mostly move the loop in opposite directions, so the deviation is plausible but unconfirmed in detail. (F, C, N)

60-65: Rosetta's loop deviates at the heterodimer interface. All X-ray homologs match 1UGH despite some not having a heterodimer partner or having different space groups. We must assume the deviation is an error. (F, M, R)

74-79: Strong crystal contact. The Rosetta ensemble and the ensemble of X-ray structures exhibit similar degrees of flexibility for this loop but mostly move it in opposite directions. (F, C, N)

isolated monomer simulation

crystal lattice simulation

1UNP - pleckstrin homology domain of protein kinase B (see also Fig. S2g)

3-6: Rosetta chooses 2 distinct N-terminus alternates. X-ray structures do not confirm, but either have native crystal contact or bind a ligand in the crystal. (F/V, C/L?, N)

15-20: Rosetta thinks loop with crystal contacts is flexible. X-ray structures disagree with each other here. Isologs 1H10, 2UVM and 2UZS are ligand-bound and more closely resemble Rosetta's alternate. (F, C/L?, S)

40-51: Rosetta thinks loop with crystal contacts is flexible. X-ray structures with slightly different crystal contacts disagree with other but do not exactly match Rosetta's alternate. Flexibility plausible though. (F, C, N)

78-83: Loop with crystal contacts. Rosetta's motion is well spanned by X-ray structures 1UNR to 2UVM/1H10. (F, C, S)

110-121: Cannot confirm Rosetta's flexibility for extended C-terminal helix with crystal contacts; all X-ray structures are close to native. (F, C, N)

isolated monomer simulation

crystal lattice simulation

1URN - RNA-binding spliceosomal protein + RNA hairpin (see also Fig. 4)

46-53: Models adopt helical backbone instead of native RNA-binding loop. One chain of X-ray isolog 1NU4 matches Rosetta; the other chain matches native. These two discrete backbone conformations seem approximately isoenergetic! (V, L, S)

78-79: Models slightly shift loop in crystal contact (within asymmetric unit). X-ray homologs are very close to native. NMR homolog 1AUD differs from native but doesn't match Rosetta's models. (V, C, N)

isolated monomer simulation

crystal lattice simulation

1UTG - oxidized uteroglobin (see also Fig. 3c)

1-17: Rosetta moves whole first helix, with crystal contacts at N-terminal end but no dimer contacts. No support from X-ray structure 2UTG, but that has same lattice as 1UTG. Seems somewhat reasonable. (F, C, N)

30-45: Rosetta models helix as flexible. X-ray structure 2UTG has same dimer and crystal contacts that influence native, so cannot confirm Rosetta's floppy alternate. (F, C/M, N)

55-70: Rosetta models C-terminal helix as flexible. X-ray structure 2UTG has same dimer and crystal contacts that influence native, so cannot confirm Rosetta's floppy alternate. (F, C/M, N)

isolated monomer simulation

crystal lattice simulation

1VCC - N-terminal domain of Vaccinia virus PA topoisomerase I

73-77: Terminus is very slightly disordered, with no confirmation. Full topoisomerase extends terminus as helix (e.g. in homologs), so flexibility is plausible for this truncated single-domain construct. (F, _, N)

Otherwise, Rosetta models and DNA-bound homologs (double mutants) are very close to deposited structure. Implies that DNA binding does not significantly alter this structure.

isolated monomer simulation

crystal lattice simulation

1VIE - R67 dihydrofolate reductase

Overall, very close to deposited structure.

34-38: Loop is very mildly mobile, almost certainly due to lack of tetramer partners. Iso/homologs match deposited structure, but are also tetrameric. (V, M, N)

44-51: Loop is more mobile, almost certainly due to lack of tetramer partners. Iso/homologs match deposited structure, but are also tetrameric. Trp45 is incorrect as deposited (MolProbity rotamer outlier, Cbeta deviation, steric clash), but many models adopt valid p-90 rotamer, corroborated by distal point mutant (Q67H) 2P4T. (F, M, N)

isolated monomer simulation

crystal lattice simulation

1VKK - mouse Glia maturation factor gamma (+sc) (see also Other Examples)

74-84: Loop flails in absence of crystal contact. NMR isolog and 81% identical NMR homolog both converge to essentially the same conformation. The Rosetta swath contains that conformation, but is probably too broad to be reasonable. Would be interesting to see if Rosetta's models violate NOE constraints or not, however. (F, C, R)

124-142: C-terminus (turn-helix) does not settle down in Rosetta models. Crystal contact would seem to make deposited structure artificially rigid, but NMR isolog and homolog are disordered only at very end of helix. More sampling is probably needed to reach convergence. (F, C, R)

isolated monomer simulation

crystal lattice simulation

Sidechains:

41: Models and MolProbity flip Gln rotamer to represent crystal-contact-free conformer, although models' backbone is somewhat floppy as well. (V, C/E, S)

55: Models and NMR isolog agree on different chi1 (but floppy chi2-3) for lattice-stabilized Glu. (F/V, C, S)

87: Crystallographic Cys alternate A clashes. Models better match non-clashing alternate B, but NMR isolog matches alternate A (albeit with a backbone shift). (F, E, N)

96: Crystallographic Cys alternate B clashes. Models nevertheless match alternate B, whereas NMR isolog matches non-clashing alternate A. (F, E, R)

97: Obvious Lys crystal contact. Models converge on alternate, but NMR isolog converges on different alternate. (V, C, N)

101: Models tuck Gln sidechain into area occupied by two well-ordered (B<25) waters; interestingly it places its terminal N and O atoms right on top of the water O atoms. NMR isolog more closely matches deposited because waters, though invisible, presumably influence the rotamer in solution. (V, W, R)

9 & 110: Obvious Glu & Arg crystal contacts. Models adopt floppy alternates but NMR isolog adopts different floppy alternates. (F, C, N)

1VLS - aspartate receptor periplasmic domain (see also Other Examples)

Generally disordered. Lack of dimer partner may play some role, but more sampling is probably needed.

75-85: Many models move this loop more so than other regions of the protein. Isologs are also extremely variable here, particularly those with aspartate bound in nearby sites. ones. Seems to be related to aspartate binding, though this loop is also part of the dimer interface. (F, C/L, N)

150-180: Some models shift C-terminal helix (preceding membrane in vivo). Suggestive of hypothesized "helix piston" motion for mechanical, transmembrane transmission of information. However, shift is more sideways than vertical, though that could be because the transmembrane region is not simulated. At least thought-provoking! (F, _, N)

isolated monomer simulation

crystal lattice simulation

1WD6 - E. coli JW1657 (see also Fig. 9)

78-82: Strand firmly continues to pair up with adjacent strand (2-10), forming idealized sheet longer than it should. Unrelated to crystal or dimer contacts. In deposited structure and homolog 2HIQ (with only a distal one-residue insertion), an ordered water (B<15) wedges between the strands here, peeling apart the sheet and occluding Rosetta's alternate. Clear-cut example of deficiencies of implicit solvent model. (V, W, R)

88-98: N-terminal helix makes the intertwined dimer interactions and deviates most; the rest of the fold is very close to native. Both chains of dimeric NMR isolog actually differ from deposited helix but in a different way. Flexibility seems possible here, but Rosetta's alternates remain unconfirmed in detail. (F, M, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1WDV - APE2540, putative trans-editing enzyme ProX

21-31: Minor helix-loop deviation at interface between two chains in unit cell. The two chains differ most relative to each other here as well, and chain B somewhat matches Rosetta's alternate. (F/V, C, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1WHO - allergen Phl P 2

1-4: Rosetta and NMR isolog move terminus to basically the same place, occluded by crystal contact in deposited structure. 1WHO has higher B-factors here and terminus of X-ray isolog 2VXQ (different lattice) is not resolved, further confirming the flexibility seen in the Rosetta ensemble. (F, C, S)

21-30: Flexibility in loop near N-terminus and with crystal contact. NMR structure is variable, but MolProbity errors suggest that is due to lack of convergence rather than true mobility. (F, C, N)

67-77: Flexibility in loop near 21-30, in dimer interface, and with crystal contact. NMR structure is variable, but MolProbity errors suggest that is due to lack of convergence rather than true mobility. (F, C/M, N)

isolated monomer simulation

crystal lattice simulation

1WIT - immunoglobulin superfamily module of twitchin, minimized average NMR

1-93: Entire structure is somewhat disordered, even more so than the NMR ensemble 1W1U, of which 1WIT is the minimized average structure. At least some of the variability in 1WIU is likely due to uncertainty as opposed to true mobility (if NMR structures in general are any indication); thus Rosetta's flexibility in excess of that observed in 1WIU is likely due to lack of convergence rather than any reflection of true flexibility -- or at least the two coexist in the ensemble and cannot be disentangled. More sampling needed. However, would be interesting to see if Rosetta's models fit within NOE constraints. (F, _, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

1WJG - probable ATP binding protein from T. themophilus HB8

80-136: Significant deviation widespread over C-terminal half of protein. Plausible pre-dimerization states. (F, M, N)

107-119: Several distinct alternates for loop with dimer and crystal contacts. Isologs confirm flexibility is possible, but do not match any of Rosetta's alternates in detail. MolProbity rotamer/Ramachandran errors indicate difficulty in fitting this loop, further supporting existence of flexibility. Apo isolog with same crystal contact (2Z3V) matches deposited much more closely than ATP-bound isologs (2Z08, 2Z09) with different lattices, so it's difficult to say what influences this loop more: ATP binding or crystal contacts. (F, C/L/M, N)

isolated monomer simulation

crystal lattice simulation

1WOU - thioredoxin-related protein 14 (see also Fig. S2h)

30-40: 80% identical NMR homolog 1V9W closely matches Rosetta's considerably different conformation for this crystal-contact-influenced loop, but no 100% homologs are available to additionally confirm this change. (F, C, S)

75-90: Cannot confirm Rosetta's helix "smear" at crystal contact. (F, C, N)

99-101: Rosetta hinges loop by up to 3Å. 80% identical NMR homolog 1V9W has locally native sequence and adopts loop alternate about 2.5Å from native, thereby confirming Rosetta's flexibility. (F, C, S)

isolated monomer simulation

crystal lattice simulation

1XD6 - gastrodianin, mannose-specific lectin (+sc)

86-94: Loop mostly converges to different position. Homolog 1XD5 matches 1XD6, but also has crystal contact (though different from 1XD6). Plausible but cannot confirm. (F/V, C, N)

105-110: C-terminal tail is well-ordered in deposited position (as indicated by clear density and low B-factors), but plausibly disordered when perturbed by adjacent 86-94 deviation. (F, C, N)

isolated monomer simulation

crystal lattice simulation

Sidechains:

22: Models and homolog (with different local crystal contact from deposited) adopt alternative Leu rotamer at crystal contact. (V, C, S)

30: Models and MolProbity flip Asn 180 degrees to alleviate deposited clash. (V, E, S)

46: Models differ from deposited Asn rotamer at crystal contact. Homolog 1XD5 lacks deposited crystal contact but strangely still matches deposited plus-chi1 rotamer (because by same authors?). (V, C/W, N)

75: Models and MolProbity flip Asn 180 degrees, but models also slightly alter chi2 and move into crystal contact to alleviate clash. Homolog 1XD5 matches deposited because has same crystal contact. (V, C/E, N)

77: Models and MolProbity flip Asn 180 degrees, but models also alter chi2 and move into crystal contact to alleviate clash. (F, C/E, N)

89: Models differ from deposited Arg rotamer at crystal contact. Homolog has different lattice from deposited locally and adopts similar rotamer to models. (F, C, S)

99: Models adopt different Ser rotamer from deposited and homolog, which both have water-mediated crystal contacts (but different from each other). (V, C/W, N)

1YNV - RNase Sa

39-42: Majority of models converge on altered loop conformation because crystal contact is not modeled. NMR structure is pretty close to Rosetta! (V, C, S)

58-65: Rosetta thinks long loop in crystal contact is very flexible. The set of crystal structures and the NMR structure, all with distant, uninvolved point mutants, are also heterogeneous but less so. (F, C, N)

75-77: Rosetta models deviate at loop with crystal contacts. Many X-ray structures with different, non-native crystal contacts match Rosetta, but they also mutate the nearby Lys94 (which interacts with the loop) to Gln (which does not), so Rosetta thinks this mutation-coupled backbone change is rather low-energy! (V, C/?, S)

isolated monomer simulation

crystal lattice simulation

256B - cytochrome b562

1-10: N-terminal helix collapses inward to heme site. However, apo NMR structures instead simply have disordered helix with no net shift. (F, L, R)

42-54: Strong crystal contacts seem to hold loop down, packed against helix ends and edge of heme. Apo (and holo) NMR structures display variability and move toward the crystal contact, whereas Rosetta moves away from it. Those structures have MolProbity errors, but they are not necessarily localized to this loop. It is safest to assume that the deviation is due to a force field deficiency rather than representing a true in-solution or apo form. (F, C/L, R)

94-106: C-terminal helix collapses inward to heme site. However, apo NMR structures instead simply have disordered helix with no net shift. (F, L/M, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

2ACY - acylphosphatase (ACN)

Generally very good agreement. Deep energy funnel.

isolated monomer simulation

crystal lattice simulation

2APB - T cell receptor Vbeta 8.2 domain

1-6: Terminus varies when crystal contact is not simulated. Ensemble of homolog crystal structures shows similar disorder, though NMR homolog is not much help. (F, C, S)

35-46: Loop appears to be flexible. Ensemble of close X-ray homologs (95+% identical) supports the deviation by matching Rosetta's swath well. Close NMR homolog is also variable here, though exact match is not quite as good. (F, C, S)

94-108: Very similar to 35-46. (F, C, S)

isolated monomer simulation

crystal lattice simulation

2CBM - neocarzinostatin after directed evolution to bind testosterone (+sc)

38-46: Loop forming ligand binding site collapses inward. 94-100% identical X-ray structures disagree with one another in a manner reasonably similar to Rosetta's models. Plausible apo conformation. (F, L, S)

95-108: Loop forming ligand binding site collapses inward. However, 94-100% identical X-ray structures, both holo and apo, match the deposited loop, discrediting the alternate. 94% identical NMR structures differ wildly and are difficult to interpret. Safest to assume a Rosetta error. (F, L, R)

isolated monomer simulation

crystal lattice simulation

Sidechains:

71: Surface-exposed deposited Arg rotamer clashes. Models prefer one alternative but are rather floppy. Homologs are also floppy but mostly populate other rotamers. (F, E, N)

74 & 82: Models and homologs have similarly floppy Glu & Arg relative to lattice-stabilized deposited conformers. (F, C, S)

89: Surface-exposed deposited Thr rotamer clashes. Some homologs match one of two model alternates. (F, E, S)

98: Deposited binding pocket Ser interacts with ligand. About half of models adopt alternative. Homologs are either too floppy or match deposited. (F, L, N)

2CCV - helix pomatia agglutinin

Rosetta models envelop crystal structure with no deviations, and qualitatively agree with B-factors. Suggests that each monomer is essentially pre-formed prior to oligomerization and GalNAc binding.

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

2CHF - Mg(2+)-unbound CheY, chemotaxis regulator (+sc)

49-52 & 76-80: Very small deviation unrelated to magnesium binding. Ordered waters (which Rosetta does not model) seem to be involved. (V, W, R)

2-4 & 127-129: Rosetta slightly shifts interacting N- and C-termini. Isologs match deposited structure, although the same crystal contacts may be in play. Most likely errors in the details of the Rosetta energy function. (F, C, R)

isolated monomer simulation

crystal lattice simulation

Sidechains:

5: Models mostly agree on alternative Glu rotamer at water-mediated crystal contact. Homologs match deposited because they also have water-mediated crystal contacts. (F, C/W, N)

15: Models adopt plausible alternatives to clashing Ser rotamer outlier; some homologs match. (F, E, S)

17: Models mostly agree on alternative Met rotamer at crystal contact. Homologs match deposited because they also have explicit water on one side and crystal contact on the other. (V, C/W, N)

19: Models adopt plausible alternatives to clashing Arg rotamer outlier, but no homologs match. (V, E, N)

26: Models adopt two distinct alternatives to Lys rotamer outlier. One of these would contact the crystal lattice and is matched by many homologs with different lattices from deposited. (F, C/E, S)

28: Models and many homologs flip deposited Leu rotamer outlier 180 degrees. (V, E, S)

44: Models and MolProbity flip Asn 180 degrees, despite possible influence of crystal contact (clash is internal). (V, C/E, S)

81: Models and many homologs flip deposited Leu rotamer outlier 180 degrees. (F, E, S)

106: Models and most homologs adopt alternative to clashing deposited Tyr rotamer outlier. Alternative moves toward ordered water and away from stabilizing, water-mediated crystal contact. (V, C/E/W, S)

2CI2 - serine proteinase inhibitor CI-2 from barley seeds

19-83: Entire structure does not really converge. Probably because other members of hexamer are not modeled. Crystal contacts likely do not play a significant role since computed models differ amongst themselves much more than NMR point mutant (E61Q) 3CI2 differs within ensemble. Rosetta error. (F, M, R)

isolated monomer simulation

crystal lattice simulation

2CMX - PA-binding winged-helix protein F-112

66-75: Loop moves in onto protein. Weak electron density suggests such hinge-like mobility is possible. Plausible, but no confirmation since the two available structures are identical. (V, _, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

2D4F - human beta2-microglobulin (see also Fig. 7, S2i)

13-21: Rosetta places loop close to protein rather than extending away from protein as in 2D4F, where it makes crystal contacts. NMR structure (1JNJ) and X-ray structure with different lattice (1A9B) match Rosetta perfectly! (V, C, S)

46-51: Loop is slightly disordered in Rosetta ensemble and NMR ensemble 1JNJ. (F, C, N)

isolated monomer simulation

crystal lattice simulation

2FLS - human glutaredoxin 2 + glutathione

35-38: For native-like subset of low-energy models, loop deviates at glutathione binding site. 90% sequence identity X-ray homolog matches native despite different crystal contacts because it is also bound, but apo 96% sequence identity NMR homolog matches direction of Rosetta's loop swing. (V, L, S)

103-115: For native-like subset of low-energy models, C-terminal helix is floppier than indicated by X-ray/NMR homologs. May be due to Rosetta's failure to place the 28-113 disulfide which should tie down the helix, or maybe more sampling is needed (F, ?, R)

isolated monomer simulation

crystal lattice simulation

2H28 - E. coli YeeU

16-20: N-terminus diverges in absence of dimer-like lattice contact. Chain B matches chain A because the contact is symmetrical, and homologs have divergent sequences, so difficult to confirm. (F, C/M, N)

118-124: C-terminus appears even more disordered than the N-terminus. Deposited chain A has crystal contact and is visible in electron density through 124, but chain B lacks crystal contact and is invisible past 118, presumably due to disorder! Sufficient evidence to support Rosetta's alternate swath. (F, C, S)

isolated monomer simulation

crystal lattice simulation

2HBO - unknown thioesterase superfamily protein from C. crescentus

Not converged. More sampling required.

isolated monomer simulation

crystal lattice simulation

2HE4 - second PDZ domain of human NHERF-2 (SLC9A3R2)

Several deviations are plausible, but all homolog structures are around 70% identity -- not enough to validate.

145-149: N-terminus flails in swath occluded by crystal contacts. (F, C, N)

167-174: Loop moves further into the protein. (F, C, N)

229-234: C-terminal loop forms strand with symmetry group residues 163-167. (F, C, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

2HH6 - B. halodurans BH3980 (see also Fig. 3d)

1-20: Rosetta's N-terminal helix is disordered. Native crystal contacts bury significant hydrophobic surface area. Rosetta's deviations seem reasonable, i.e. helix could be stabilized by crystal contacts and flexible in solution, but no homologs exist to confirm. (F, C, N)

81-91: Rosetta's helix at dimer interface is significantly disordered. No homologs exist to confirm. (F, M, N)

100-121: Half of Rosetta models match deposited position for last helix, deeply in dimer interface (despite not modeling dimer partner); other half tuck it into the body of the protein. These wildly different but distinct alternatives suggest that dimer formation may not alter the energetic landscape of the monomer so much as select out one of its minima. (F, M, N)

isolated monomer simulation

crystal lattice simulation

2HL7 - periplasmic domain of P. aeruginosa CcmH

-4-6: N-terminus adopts long, disperse alpha helix instead of short 310 helix preceded by tail in extended (beta) conformation. Cannot confirm or refute due to lack of close homologs. Tail is probably crystal-contact-induced so computed models are reasonable for -4-1, but 310 instead of alpha helix for 2-6 is more likely a modeling error. (F/V, C, N/R)

75-78: C-terminus flails in Rosetta models. Deposited structure has no crystal contact but does have high B-factors and unclear electron density, suggesting flexibility is likely, but cannot confirm due to lack of close homologs. (F, _, N)

isolated monomer simulation

crystal lattice simulation

2HNG - unknown S. pneumoniae protein SP1558

85-95: Segment composed of 310 helix, turn, alpha helix is somewhat flexible. 310 i,i+3 mainchain H-bond and Asn92 canonical N-cap H-bond do not form. Lack of explicit water (B<39) from deposited structure seems to impede convergence. Good electron density and no MolProbity errors, but deposited crystallographic alternate conformations for 86-87 and 90. Possibly crystal-contact-induced, but intuitively seems more likely to be a Rosetta force field problem. (F/V, C/W, R)

120-128: Dimerization occludes the alternative binding site for the C-terminus: extra strand in central sheet instead of helix on other side of protein. Plausible but cannot truly confirm. (F/V, M, N)

isolated monomer simulation

crystal lattice simulation

2HSB - unknown A. fulgidus protein from DUF103 family

Considerably divergent conformation globally. More sampling may be needed, though this protein may simply be somewhat disordered before dimerizing...

83-96: Particularly large deviation at helix-turn comprising dimer interface. A plausible but unconfirmed pre-dimerization state. (F, M, N)

isolated monomer simulation

crystal lattice simulation

2HSH - C73S mutant of human thioredoxin-1 oxidized with H2O2 (+sc)

7-20: Helix is shifted in models, but X-ray homologs match native instead. Probably because Rosetta doesn't model salt bridges like Lys8-Glu68 at native helix N-terminus. (F, ?, R)

35-49: Rosetta has both correct and incorrect helix conformations at about equal energy. SS bridge at helix N-terminus probably not relevant because both conformations can form a decent SS bridge. However, incorrect conformation would clash with some crystallographic waters, which could be the problem. (V, W?, R)

71-73: Low-energy models swing loop away from protein core with ordered water toward crystal contact. Some homologs deviate in the same direction. Native actually has 72-73 crystallographic alternate backbone! (Incidentally, both deposited Lys72 rotamers are MolProbity outliers.) (F, C, S)

isolated monomer simulation

crystal lattice simulation

Sidechains:

4: Models choose different Gln rotamer from deposited and homologs with same lattice. (V, C, N)

43: Models choose different His rotamer. Some homologs (1AUC, 2IIY) somewhat agree with models. (V, S)

68: Models choose wrong Asp rotamer because Rosetta does not model the Ser7-Asp20 salt bridge. (F, R)

72: Models and homologs agree on alternative to Lys rotamer outlier. (V, E, S)

90: Deposited structure and homologs have 2/3 chi1 rotamers. Models sample all 3/3. (F, N)

2I1U - M. tuberculosis thioredoxin C

7-10: N-terminal tail collapses inward, away from crystal contact. (F/V, C, N)

106-114: C-terminal helix continues instead of straightening out in sheet-like interaction with lattice neighbor. (F/V, C, N)

(Some other non-convergence may be resolved by more sampling.)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

2IB0 - conserved M. tuberculosis hypothetical protein rv2844

86-91: This loop in deposited chain A is "pulled" by a crystal contact. Rosetta's models depart and could more closely reflect the equilibrium conformation in solution. There are no homolog structures, but deposited chain B differs most from chain A here. The difference is in the opposite direction from Rosetta, but that may be because it also has a crystal contact. Somewhat higher B-factors here as well. A degree of inherent local plasticity is extremely likely, though Rosetta's exact alternate cannot be confirmed in atomic detail. (V, C, N)

151-158: C-terminus is disordered, which is extremely reasonable given its lack of interactions with the main body of the protein. Note that electron density is essentially non-existent for deposited chain B, further validating the presence of disorder. (F, C, S)

(Dimerization occurs via helix-helix packing, and appears to involve no appreciable conformational change.)

isolated monomer simulation

crystal lattice simulation

2ICP - bacterial antitoxin HigA

74-94: C-terminal helix folds in on itself and flails a bit. Isolog 2ICT demonstrates the possibility of flexibility, although it does not necessarily agree with the Rosetta deviation -- it is also in a crystal. (F, C, N)

isolated monomer simulation

crystal lattice simulation

2O7K - S. aureus thioredoxin

Generally not quite converged, probably due to insufficient sampling.

8-18: Helix is excessively disordered, as confirmed by tight ensembles formed by crystal and NMR homologs. (F, _, R)

isolated monomer simulation

crystal lattice simulation

2VIK - villin 14T, minimized average NMR

Generally not quite converged -- probably needs more sampling.

1-15 & 116-126: N- & C-termini are especially disordered, more than suggested by NMR ensemble 2VIL (of which 2VIK is the minimized average structure) and 61-62% X-ray homologs. Likely erroneous. (F, _, R)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

4ICB - calbindin D9k

0-75: Several low-energy models are globally wildly different from deposited structure -- disregarded for specific converged deviations below -- but the majority match it quite well. The minority's disorder may be due to lack of calcium atoms in Rosetta simulation, but if so it is surprising that the essentially native-like structure is the dominant conformation -- perhaps there is a shallow energy basin around a sharp energy funnel. Another possibility is simply that more sampling is needed. (F, L, R)

0-8: Crystal contact seems to pull N-terminus away from Rosetta's equilibrium position. Distal from calcium-binding sites. X-ray isologs (with different lattice from 4ICB) and NMR homologs vary here, confirming plasticity. (V, C, S)

37-47: Small, seemingly crystal-contact-induced deviation for slightly irregular helix-turn-helix. Distal from calcium-binding sites. MolProbity reveals that Lys41 is misfit as deposited; perhaps it was difficult to fit due to weak density stemming from mobility, but structure factors are unavailable to confirm this idea. The region is deposited with extensive mainchain alternate conformations indicative of mobility. The variability of X-ray isologs (with different lattice from 4ICB) and NMR homologs further confirms plasticity. (V, C/E, S)

68-75: Crystal contact seems to push end of C-terminal helix away from Rosetta's equilibrium position. Distal from calcium-binding sites. X-ray isologs (with different lattice from 4ICB) and NMR homologs vary here, confirming plasticity. (V, C, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

4UBP - B. pasteurii urease + acetohydroxamate anion

1-7: N-terminus folds to body of protein, very obviously because hetero-oligomer interaction is not modeled. No corroborating evidence, but visually the reason for the deviation is absolutely clear-cut. Enough to be convincing as a pre-oligomerization state. (F/V, M, S)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

5CRO - Cro repressor protein from bacteriophage lambda

1-5: N-terminus is mobile, but normally interacts within the tetrametric structure. Very likely due to single subunit simulation. Plausible as a hypothetical transient monomer prior to tetramerization. (F, M, N)

43-51: Hairpin adjacent to N-terminus. Similar analysis. (F, M, N)

isolated monomer simulation

crystal lattice simulation

(No simulation in crystal lattice)

Legend for Deviation Descriptions

Format used above: "Description of deviation. (Flexibility, Cause, Support)"

Flexibility

F = floppy

V = converged

Cause

C = crystal contact

M = multimer

L = ligand

W = water

E = error in deposited structure

Support

S = supported

N = not supported but plausible

R = Rosetta error

Return to top of page

Click a figure below to view an interactive 3D kinemage! (Java required)

Rosetta Error: 1TTZ - genomics target X. campestris XcR50

Rosetta compromises rotamer quality to move helix

Rosetta Error: 1VKK - mouse Glia maturation factor gamma

Rosetta places Gln OE1, NE2 where ordered waters should be

Allostery: 1VLS - aspartate receptor periplasmic domain

Rosetta lowest-energy models are most variable at pre-membrane helix "piston"