AlphaFold
RoseTTAFold
Exercises
Solutions

HPC: Analyze the predicted SarsCoV2-VHH.E complex

You should have already downloaded or generated AlphaFold predictions in the previous exercises. The visualization plots for the SarsCoV2-VHH.E complex can be found here (PLDDT) and here (PAE). In this exercise, we will go through the predictions and visualizations, in order to interpret them appropriately.

  1. Model rankings. Determine which of the predicted structures were predicted with highest confidence. You should find a significant gap between the best model and the others. Hint: ranking_debug.json holds useful information on the predicted TM scores, which represent the global confidence of the protein complex.
  2. PLDDT. Check in which regions of the complex the local confidence is high, for the top ranked model and any of the other models. The local confidence, given by a predicted LDDT score per residue, can be found in two ways.
    1. In the prefix_coverage_LDDT.png file
    2. In the PDB files, it is stored in the b-factors column. Use the YASARA / PyMol software to color residues by b-factors, which will give you a visual overview of the PLDDT.
  3. PAE. The predicted aligned error is given in the prefix_PAE.png file. Inspect the file for this complex. Are there models where AlphaFold is fully confident of the prediction of the complex? Can you link the visualization to the PTM scores that you inspected in step 1.?
  4. Aligning predicted structures to 7KN5. As a final validation step for the predictions of this complex, we can compare them to the experimentally determined 7KN5 structure as found in PDB (download link for a slimmed-down version with just one RBD and one VHH.E). First, align the top ranked model with the lower ranked models. Do you see differences? Next, align the top model to the 7KN5 structure. Do you find a good match? Try to align the ranked models with low confidence as well. What do you see? Does this confirm what you saw in the predicted TM score ranking and the PAE plots?