Fig 1.
The Structure view is a window into a predicted structure and a host of tools and associated information.
Fig 2.
A. Structure Panel: the left panel in the Structure View displays the structure. This is the predicted structure of a homodimer of the polyketide synthase pik2, with NADPH docked into the AT domain with AutoDock Vina. This structure was rendered with NGLViewer [15]. B. Sequence View: the amino acid sequence is displayed in the Inputs tab, and is optionally annotated with domain predictions, such as the pfam annotations displayed here. C. Predicted Alignment Error: The predicted alignment error, measured in Ångstroms, is an AlphaFold metric. In Foldy it is rendered in the PAE tab with Plotly [16]. This displayed heatmap shows the predicted alignment error for the pik2 homodimer. The relatively low PAE between the KS-AT portions of the two peptides indicates that the two KS-ATs rigidly dimerize, while the comparably high PAE between the KRs and the rest of the structure indicate that the KRs move freely and independently. D. Contact Probability: The contact probability is derived from the internal AlphaFold distance probability distribution function, and can be interpreted as the likelihood that two residues are within 8Å of each other. For residues between disparate chains, this might suggest a contact point between two interacting domains or peptides. The 8Å contact probability is rendered in Foldy on the Contacts tab with Plotly [16]. Displayed is the contact probability for the predicted structure of the pik2 homodimer. Yellow dots in off-diagonal blocks suggest predicted interacting or contacting residues between the two peptides.
Fig 3.
A fatty acyl-AMP ligase with substrate docked.
The structure of the wild type FadD23 from M. smegmatis as predicted by AlphaFold in Foldy is shown in blue, rendered in pymol. Within the active site is the pose of octanoyl-AMP docked as predicted by AutoDock Vina in Foldy. This pose was chosen from the ensemble of predicted poses based on biochemical information about the active site. Two notable residues are highlighted: G321 and L218. The wild type backbone and side chain are displayed for both in blue. Displayed in orange and red are the same residues taken from two different structures which are mutants of the wild type. On top is residue 321 from the G321W mutant’s structure as predicted by AlphaFold. This mutant seems to occlude the proper insertion of the C8 ligand. Below is residue 218 from the L218W mutant’s structure as predicted by AlphaFold. This mutation seems to allow insertion of the C8 ligand but would occlude insertion of longer ligands like the C10-AMP. Based on these three structures, L218W is a good candidate for making an FAAL which prefers shorter chain fatty acyl AMPs while preserving activity for the C8 chains.
Fig 4.
Putative complex formation of two domains of unknown function in P. putida.
Four genes known to be essential for metabolism of Tween-20 were co-folded with AlphaFold in Foldy. A. Peptide-Peptide Contact Probability: As done in Humphreys et al [6] Foldy approximates the peptide-peptide contact probability as the maximum contact probability of any residues between two cofolded peptides. The matrix of peptide-peptide contact probabilities is shown in the Contact tab of the Structure View page. This contact probability matrix shows that maybe: PP_0766 and PP_2019 weakly interact, PP_0766 and PP_0765 weakly interact, and PP_2019 and PP_2018 strongly interact. B. Predicted Interactions of Phenotypically Related Proteins: The four proteins whose knockouts have related phenotypes were folded in Foldy, and although all four cluster closely, the peptide-peptide contact probability map indicates that not all four interact directly with each other. By using biochemical knowledge about these proteins, we are led to suspect that there are two different complexes which can form. C. Novel Hypothesized Substrate Transport System: Two DUFs in P. putida, previously hypothesized to have hydrolase activity may actually be involved in substrate transport: PP_0765 (DUF1302, top blue) and PP_0766 (DUF1329, top green, bottom green). PP_0765 has the characteristic beta-barrel of a membrane protein, and shows high likelihood of forming a complex with PP_0766 (top). Additionally, PP_0766 is predicted to form a heterotrimer with PP_2018 (bottom blue) and PP_2019 (bottom red).
Table 1.
Comparison of different structure prediction tools.