9. Morph the predicted model onto the docked fragments and rebuild the missing parts of the model in the map density
In this section of the AlphaFold tutorial we are going to rebuild the docked isoform CRA_a of TACAN monomer structure finally obtained in the previous section.
We are going to take advantage of the protocol phenix-rebuild predicted model to rebuild the docked structure prediction into the map (protocol appendix). As the title of this section indicates, the rebuild function morphs the predicted model onto the docked fragments and rebuilds the missing parts of the model in the map density. The result, however, depends very much on the map we are working with. In any case, the result can still be improved by additional rounds of processing.
Let us start by opening (Fig. 9.24 (1)) and filling in the protocol form, adding the previously obtained predicted AlphaFold2 (2) and docked models (3), and the map (4). After fixing the map resolution limit, increase the number of threads (5) and finally, execute the protocol (6).
After a quick running, the protocol finishes. Probably due to a file format incompatibility, the tool was unable to morph the predicted structure, as you can check in the results clicking
Analyze Results. Nevertheless, we can use the recently obtained docked model (Fig. 8.13) as template to get a new AlphaFold prediction using this option in the Phenix Colab Notebook. With this aim, complete again the chimerax-alphafold prediction protocol (Fig. 9.25 (1)) selecting Google Colab (2), the Phenix Colab Notebook (3), the isoform CRA_a of TACAN protein sequence (4) and the previous docked model (5). Then, execute the protocol (6).
Soon after executing the protocol, the browser window of Phenix Colab Notebok opens and stars to run.
REMARK that when a template is used, this template has to be manually uploaded. This task requires clicking the key
Choose Files in the Cell #3 of the browser (Fig. 9.26 (1)).
After a while the prediction process finishes and the AlphaFold2 predicted structure is shown in ChimeraX. Fig. 9.27 details the new prediction (A) and the PAE plot (B). Comparing the accuracy of residues and PAE error with the predictions of monomers and dimers that we have obtained before, we can observe that the new one is the best one.
At this point we can follow the common workflow to process, dock and rebuild the new AlphaFold2 prediction. Starting with the model processing, open the protocol phenix-process predicted model (Fig. 9.28 (1)) with the predicted model (2), the B-value column values to consider as minimum to process (3), the PAE file (4), and then execute the protocol (5).
After executing the protocol press
Analyze Results to check the results menu (Fig. 9.29 (A)).
Clicking Structures in ChimeraX (Fig. 9.29 (A, 1)), the predicted and the processed structures will be simultaneously opened. Whereas the processed structure is still colored according to the AlphaFold2 confidence score, the processed structure is pink colored (Fig. 9.29 (B)). Remark that this processed structure is constituted by only one domain (chain) (model #3). Concerning the remaining fragments of sequence removed, open Remaining Sequences in the results’ menu (Fig. 9.29 (A, 2)) to observe the text file (C). The sequence is quite similar to the sequence shown in Fig. 7.18 (C)).
Next, run the protocol phenix-dock predicted model to dock the structure prediction into the map (protocol appendix). Open the protocol (Fig. 9.30 (1)) and complete it with the new AlphaFold2 prediction obtained with Phenix notebook (2) (Fig. 9.27). Include the respective processed prediction (3) (Fig. 9.29 (B)). Fill in the Input map box with the recently downloaded map (4) (Fig. 8.11). Update the resolution value of the map, increase the number of processors to speed up the process (5) and, finally, execute the protocol (6).
Visualize the results clicking
Analyze Results when the execution finishes. The ChimeraX GUI window will open to show the map and several models (Fig. 9.31). Not shown in Fig. 9.31, models #3 and #4 are the new AlphaFold2 prediction and the processed models, respectively. Pink model #5 is the docked model.
Select the appropriate orientation of the map, the density threshold and the appearance as mesh to better visualize the docked structure.
At this point we take advantage of the protocol phenix-rebuild predicted model to rebuild the docked structure prediction into the map (protocol appendix). Rebuild function morphs the predicted model onto the docked fragments and rebuilds the missing parts of the model in the map density. Then, open the protocol (Fig. 9.24 (1)) and fill in the protocol form, adding the AlphaFold2 prediction previously obtained (Fig. 9.27) (2), the docked model (Fig. 9.31) (3) and the map (4). After fixing the map resolution limit, increase the number of threads (5) and finally, execute the protocol (6).
The protocol finishes after a while and clicking
Analyze Results results can be checked (Fig. 9.33).
The two different views shown in Fig. 9.33 detail the morphing both of horizontal alpha-helices (A, 1) and vertical ones (B, 2). Remark that the new rebuilt model (pink) fits better to the map than the previous docked one (green). The rest of the sequence, however, was unable to adapt its architecture to the map and remains mainly disordered. This result makes sense due to the sequence differences in the C-terminal part of the protein between the two isoforms of the human TACAN protein. Only the sequence identical in both isoforms has been predicted by Alphafold2 and finally morphed according to the map of the first isoform. The rest of the structure of the second isoform remains unpredicted and seems to adopt a quite different architecture than the first isoform. The experimental map of this second isoform CRA_a is thus essential to characterize the structure of the C-terminal part of the second isoform.