Joining AlphaFold predictions for halves of a molecule

From Proteopedia

This page is a supplement to How to predict structures with AlphaFold. First, be sure to find out if the AlphaFold Database already has a prediction for your sequence. If your query sequence is >1,000 amino acids, AlphaFold Colab may fail. In that case, you can try making predictions for two overlapping halves of the sequence, and then joining the two predictions into one.

For a bit more detail of the method, see the footnote that begins "I had one sequence of length ~1,300." at the bottom of How to predict structures with AlphaFold.

For superposing the overlapping segments, DeepView is a powerful option, but it is not user friendly. An easier method may be the SuperPose Server. It enables you to specify the overlapping sequence ranges of the two structure predictions so the superposition will focus on the overlap.

For cutting and pasting PDB files, be sure to use a Plain text editor.

After the PDB files have been joined, you may wish to renumber the sequence. See How to renumber PDB files.

Here are simplified sketches to summarize the method:

See also AlphaFold/Index, a list of pages in Proteopedia about Alphafold.