Predicting initial models with AlphaFold
AlphaFold neural network predicts the 3D coordinates of all heavy atoms for a given protein using solely the primary amino acid sequence as input [Jumper et al., 2021].
In this tutorial we show how to generate AlphaFold models of your sequence and rebuild them using the 3D density map. We use the Scipion framework and the approach described in the PHENIX web site and summarized as:
Get the initial AlphaFold model
Remove low-confidence residues of the model and break it into compact domains
Dock the domains obtained in the previous model processing step into your cryo-EM map or unit cell
Morph the docked fragments and rebuild the whole predicted model in the map density
Contents
- 1. Revision History
- 2. Prelude
- 3. Introduction to Model building with AlphaFold
- 4. Problem to solve: Human ion channel TACAN isoform induced by tumor necrosis factor alpha
- 5. General workflow followed in this tutorial to predict and fit Alphafold2 structures
- 6. Get the initial model with AlphaFold2
- 7. Remove low-confidence residues of the model and break it into compact domains
- 8. Dock the domains obtained into your cryo-EM map or unit cell
- 9. Morph the predicted model onto the docked fragments and rebuild the missing parts of the model in the map density
- 10. Three in one: Processing, docking and rebuilding the model in the map density