Visualize Violations

This is a script for visualizing violated predictions in PyMol.

Instructions

First, clone this onto your local computer, with a command like:

git clone https://github.com/ProteinDesignLab/visualize_violations.git

Then, place the following files into the files folder, where <prot_id> is the CAMEO ID e.g. 2020-02-29_000000219_2.

• <prot_id>.pdb - This is the crystal structure's PDB
• <prot_id>_ca.pdb - This is the predicted CA trace's PDB, which can be found in the cnd.files folder under output/{database}/{folder}/cnd.files/{prot_id}, after running the PFOLD SDP+GD pipeline
• <prot_id>.npz / <prot_id>_tf_std.npz - This is the predicted distance matrix from step 3 of the PFOLD pipeline, which is can be found at output/{database}/{folder}/npz.files/{prot_id}/{prot_id}_tf_std.npz

Finally, run

pymol visualize_violations.py -- -p=<prot_id> -t=10

Note that we are running this with pymol and not python. Also note the -- between the script and the flags.

More details about the flags:

• --protein / -p - The protein CAMEO ID e.g. 2020-02-29_000000219_2
• --threshold / -t - Show only distance violations above this threshold (in Angstroms); defaults to 10

Example

The files folder already contains sample files for 2020-02-29_000000219_2, which can be viewed with

pymol visualize_violations.py -- -p=2020-02-29_000000219_2

The white cartoon is the crystal structure, while the magenta lines highlight predicted distances that are off by more than 10A, compared to the crystal structure. Here we see that there are some violations around the loop.

The yellow cartoon is the predicted CA trace from SDP+GD, while the cyan lines highlight predicted distances that are off by more than 10A, compared to the CA trace. In other words, the predicted CA trace is unable to fulfill these distance constraints.

Here, we see most of the violations stemming from the residues in the loop on the right.

The red lines represent the previous violations (CA trace vs predicted distances), superimposed onto the crystal structure for clarity. This is especially useful when the predicted CA trace is messy.

Superimposing the previous violations onto the crystal, we see that the violations coincide with the wrongly predicted loop structure, which is straight in the predicted CA trace, but pinned down in the crystal structure.