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 "cython<3.2"  # Required to avoid Cython 3.2.0 compiler bug when building cgrtools (transitive dependency with aizynthfinder)
pip install syndirella

Note: If cgrtools installation fails, try running:

conda install -c conda-forge c-compiler cxx-compiler
pip install --no-build-isolation cgrtools

Troubleshooting: If you encounter a TypeError: 'AttributeFilledMock' object is not iterable error, this is related to a PyRosetta dependency by Fragmenstein. To resolve this, you can download PyRosetta for academic and non-commercial use (see PyRosetta License).

pip install pyrosetta-installer
python -c 'import pyrosetta_installer; pyrosetta_installer.install_pyrosetta()'

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.