Authors: Andreas Tosstorff, Jason Cole, Robin Taylor, Bernd Kuhn, 2019-2023

This repository contains scripts to calculate ratio of frequencies values (RF). If you use this code or derived statistics, please cite:

• https://pubs.acs.org/doi/abs/10.1021/acs.jcim.0c00858
• https://chemistry-europe.onlinelibrary.wiley.com/doi/abs/10.1002/cmdc.202100387

Use of the RF interaction analysis as described in:

Augmenting Structure-Based Design with Experimental Protein-Ligand Interaction Data: Molecular Recognition, Interactive Visualization, and Rescoring. A. Tosstorff, J.C. Cole, R. Bartelt, B. Kuhn ChemMedChem 2021, 16(22), 3428-3438 https://doi.org/10.1002/cmdc.202100387

This can be a very powerful tool to highlight potentially favorable and unfavorable protein-ligand interactions in a binding site. However, the user should be aware of the following considerations when using the MOE RF tool:

1. RF values indicate the competitiveness for a given interaction by comparing its observed frequency of occurence against a statistical null model based on exposed surface areas. RF values > 1 indicate interactions that occur more often than expected by chance ("competitive"). Conversely, RF < 1 describes an interaction that occurs less often than expected ("non-competitive"). Non-competitive interactions are not necessarily unfavorable, it just means that other interactions are more likely to occur in the PDB. From our experience, low RF values < 0.8 likely reflect a poor protein-ligand contact and should be addressed to improve binding. The RF sliders in MOE can be used to limit the RF interaction display to the desired range.

2. Due to the implicit hydrogen-atom approach in generating the underlying statistics, the statistics is less accurate for OH groups. This applies for ex. to the sidechain oxygen atoms of Ser, Thr, and Tyr, as well as to water.

3. Short protein-ligand contacts are not well captured by the RF interaction statistics, and the MOE clash detection should be turned on to detect such cases. In particular, ligand atoms with unusually high RF scores (e.g. > 3) and many protein contacts sometimes involve a steric clash. Their interatomic distances should be checked carefully, for ex. using distance labeling in the MOE RF panel.

4. MOE RF interactions that are statistically significant are highlighted as thick lines and should be higher weighted than those with thin lines, indicating only a trend but no statistical significance.

5. In rare cases, a short protein-ligand contact might exist that is not detected by the RF interaction tool. This can arise due to a missing ligand atom type, and can be detected by checking "Ligand Type" and "Line-of-Sight" labels in the MOE RF panel.

---

Installation

conda env create -f environment.yaml -p /path/to/env
conda activate rf-statistics
poetry install

In order to get in line with current commit conventions of the team, developers can set up a **pre-commit** hook that is triggered before every commit.

To first install it is very simple:

```bash
# on the conda env
$> pip install pre-commit
$> pre-commit install
...
$> git commit -am "message" # the hooks will trigger

sHPC environment

conda activate rf-statistics3.3

Generating Databases

Binding sites should be in the Proasis format.
Proasis3, version 3.532, © 2010-2020, Desert Scientific Software, Pty Ltd, All rights reserved.

• database_utils/rdkit_library.py generates an RDKit Library for substructure search from Proasis binding sites.
• database_utils/ccdc_library.py generates .csdsql file from Proasis binding sites.
• database_utils/pdb_quality.py extracts crystal structure quality information from the PDB.

---

Calculating new RF statistics
Requires CSD API 3.0.9 or higher

• los_csv_to_lsf.py will submit jobs to the Roche LSF queue.
• Atom type definitions should be provided in CSV files. Ligand atom types should be sorted manually. Ligand atom types at the top of the list will be matched first.
• Input protein-ligand binding sites should be in the "Proasis" format, i.e. central ligand atoms start with _Z, other HET groups start with _U. Protein atoms start with _[one_letter_code].
• update_lookup_files.py writes lookup CSV files for MOE

---

Generating PyMOL scenes with RF values

# From Proasis database
rf_assignment.py -i ABCD_001
rf_assignment.py -i ABCDE_001

# From GOLD output
rf_assignment.py -i gold.sdf --gold gold.conf

---

Generating CSV with RF values for an SDF of ligands and a target protein

rf_values_from_pdb_and_sdf.py -p apo_protein.pdb -l ligands.sdf

---