xchem-syndirella 1.0.6

Syndirella (Synthesis Directed Elaborations) is a tool for generating and scoring synthetically practical elaborations of molecules designed from fragment screens. Starting from the retrosynthetic analysis of fragment merged molecules, superstructures of the original reactants are found by catalog search and filtered by selectivity issues. The elaborated final products are defined by reaction SMIRKS patterns and energy minimized in the protein with restraints to experimental data.

Preprint: https://chemrxiv.org/engage/chemrxiv/article-details/68d4f08ff416303770403a44

Documentation

Full Documentation - Complete user guide, API reference, and examples

Quick Start

Installation

conda create -n syndirella python=3.10
conda activate syndirella
pip install syndirella
pip install aizynthfinder
python -c 'import pyrosetta_installer; pyrosetta_installer.install_pyrosetta()'  # run once, for Fragmenstein

Note: PyRosetta is required by Fragmenstein to operate. Syndirella does not use PyRosetta by default to score compounds. PyRosetta is available for academic and non-commercial use (see PyRosetta License).

Troubleshooting:

cgrtools fails to install:

pip install "cython<3.2"
conda install -c conda-forge c-compiler cxx-compiler
pip install --no-build-isolation cgrtools

Install with HIPPO (optional)

pip install syndirella[hippo]
python -c 'import pyrosetta_installer; pyrosetta_installer.install_pyrosetta()'  # run once, for Fragmenstein
pip install aizynthfinder reaction-utils --no-deps # AiZynthFinder and reaction-utils declare `numpy = "<2.0.0"` and `rdkit = "^2023.9.1"` which are uncompatible with HIPPO. Install with `--no-deps` so pip does not pull the wrong NumPy and RDKit versions.
pip install tqdm paretoset networkx rdchiral wrapt wrapt_timeout_decorator xxhash swifter apted onnxruntime tables
conda install -c conda-forge chemicalite

See the HIPPO documentation for full setup and version compatibility.

Basic Usage

1. Setup AiZynthFinder (required for retrosynthesis):

   ⚠️ WARNING: This step downloads large model files (~750MB total) if run the first time. These are required to run AiZynthFinder.

   Files will be downloaded to: [syndirella_package_path]/aizynth/

   syndirella setup-aizynth
   

2. Run the pipeline:

   syndirella run --input /absolute/path/to/input.csv --output /absolute/path/to/output_directory --templates /absolute/path/to/templates/ --hits_path /absolute/path/to/fragments.sdf
   

   Note: All paths must be absolute paths.

Example Input

Create a CSV file with your fragment data. Hit names must match the exact names in your SDF file:

smiles,hit1,hit2,hit3,template,compound_set
O=C(NC1CC(C(F)(F)F)C1)c1cc2ccsc2[nH]1,A71EV2A-x0556_A_147_1_A71EV2A-x0526+A+147+1,A71EV2A-x0566_A_147_1_A71EV2A-x0526+A+147+1,,Ax0310a_apo-desolv,my_compound_set
CC(=O)Nc1cc(CC(=O)NCC(NC(=O)CCl)c2cccnc2)c(NC(C)=O)nn1,A71EV2A-x1346_A_250_1_A71EV2A-x0526+A+147+1,,,Ax1346a_apo-desolv,my_compound_set

For complete examples and templates, see the run_syndirella_example directory.

Output Files

The main output file to read is the structured_output.pkl.gz file, which contains:

• Complete synthesis routes with reaction details
• Placement information with energy values
• Structural quality metrics
• Product information and success flags

This file contains all the information needed to analyze and reproduce the elaborations.

Requirements

• Python 3.10
• RDKit
• AiZynthFinder
• Fragmenstein
• Additional dependencies (see pyproject.toml)

Learn More

• Examples - Jupyter notebooks and sample data
• User Guide - Detailed usage instructions
• Reaction Constants - Advanced reaction settings and options

License

This project is licensed under the MIT License.