Maximum Achievable Potency of a Protein Surface

Step 1: Determine Target Residue

For example: Protein Mdm2. Select 1 residue (non-buried) on the protein surface as target residue.

Or iterate over the all the protein surface residues with SASA > 10 Å

Step 2: Open pockets at the target site (Optional)

Rosetta based application that explores low-energy fluctuations of the protein surface to reveal cryptic pockets.

Sample Command: relax.linuxgccrelease -s input_pdb -relax:fast -pocket_max_spacing 12 -pocket_zero_derivatives -pocket_psp false -pocket_sps -pocket_num_angles 2 -ex1 ex1aro -ex2 -score:patch pocket.wts.patch -nstruct 1 -cst_fa_file constraints

Dependency: Rosetta Software Suite

If there is a good pocket at the target site then skip to Step 3

Step 3: Build Exemplars

“Exemplar”: a perfect, but nonphysical, pseudoligand that would optimally match the shape and chemical features of the pocket.

Dependency: Rosetta Software Suite

Sample Command: make_exemplar.linuxgccrelease -database ~/Rosetta/main/database -in:file:s input.pdb -central_relax_pdb_num 99:A -pocket_grid_size 12 -pocket_static_grid -pocket_filter_by_exemplar

Step 4.1: Obtain Pocket

Script: prox4_exemplar.pl

"Pocket": protein surface in 4Å proximity to the exemplar

Step 4.2: Calculate Pocket Features

Script: prox4_exemplar.pl && features.pl

"Pocket Features": 13 features of protein pocket geometric and physiochemical properties.

Dependencies: RADI and Naccess

------------------- Pocket Features --------------------------

| Property Name | Description | | ------------- | ------------- | | VOLUME_HULL | volume of convex hull computed using RADI software | | hydrophobicity_kyte | hydrophobicity based properties of residues | | SMALLEST_SIZE | distance separating the two closest slabs enclosing the hull computed using RADI software | | INERTIA_3 | smallest eigenvalue of inertia matrix computed using RADI software | | p_N_atom | frequency of N atoms in pocket | | hydrophobicity_pocket | hydrophobicity pocket estimated with solvent accessibility computed using NACCESS software | | p_aliphatic_residues | frequency of positive residues in pocket (I, L, V) | | p_aromatic_residues | frequency of aromatic residues in pocket (F, Y, H, W) | | SURFACE_HULL | surface of convex hull computed using RADI software | | p_negative_residues | frequency of negative residues in pocket (D, E) | | C_RESIDUE | number of residues in pocket | | p_Ooh_atom | frequency of Ooh atoms in pocket | | p_Ccoo_atom | frequency of Ccoo atoms in pocket |

Step 5: Form Set

Script: FormSets.pl

This script forms the test set.

Usage: perl FormSets.pl > DataSet

Step 6: Applying GBM model

Model Name: model_FINAL.rds

Usage: Rscript GBMrunFINAL.R

Predictions are added in the last column of the Set.

Download Prerequisites

1. Download RADI
   • Download binary from http://petitjeanmichel.free.fr/itoweb.petitjean.freeware.html#RADI
   • Change path for binary in line 32 script features.pl
2. Download Naccess
   • Download Naccess app from http://wolf.bms.umist.ac.uk/naccess/
   • Change path for the app in line 86 script features.pl

Walk-Through: Example of Mdm2

Protein: Mdm2

Target Site: Chain A Residue 99

## Skip these steps for this particular example ##

mkdir ~/PocketDruggability/data/TA99/

cd ~/PocketDruggability/data/TA99/

mkdir complexes/

cd complexes/

## Place Protein-Exemplar Complexes here. Filename: "*_Complex.pdb"

cd ../

mkdir apo/

cd apo/

## Place Protein Apo Structure files here. Filename: "*.pdb"

cd ~/PocketDruggability/

## Previous steps have already been done for this particular example. Start Here. ##

perl prox4_exemplar.pl

perl FormSets.pl > TA99Set

Rscript GBMrunFINAL.R

Output file: TA99Predictions.txt

Column15 : PredictedActivity (Predicted attainable pactivity for a pocket)

References

1. Borrel A, Regad L, Xhaard H, Petitjean M, Camproux A-C, PockDrug: A Model for Predicting Pocket Druggability That Overcomes Pocket Estimation Uncertainties. Journal of Chemical Information and Modeling 2015, 55 (4), 882-895.
2. Burgoyne NJ, Jackson RM, Predicting protein interaction sites: binding hot-spots in protein–protein and protein–ligand interfaces. Bioinformatics 2006, 22 (11), 1335-1342.
3. Petitjean M, Applications of the radius-diameter diagram to the classification of topological and geometrical shapes of chemical compounds. Journal of chemical information and computer sciences 1992, 32 (4), 331-337.
4. Kyte J, Doolittle RF, A simple method for displaying the hydropathic character of a protein. Journal of molecular biology 1982, 157 (1), 105-132.
5. Milletti F. Vulpetti A, Predicting polypharmacology by binding site similarity: from kinases to the protein universe. Journal of chemical information and modeling 2010, 50 (8), 1418-1431.