SSTR: Spectrum2Structure Transformer Ranker

The Spectrum2Structure Transformer Ranker (SSTR) is a tool designed to rank and generate chemical structures based on MS/MS spectrum data. It leverages transformers to perform de novo chemical structure generation and ranking of candidate molecules.

Installation

We recommend using conda to create a virtual environment and install the dependencies.

conda create -n SSTR python=3.10
conda activate SSTR

To install the package, run the following command:

pip install SSTR

To test the installation, run the following command:

sstr --help

If the installation is successful, you should see the help message for the SSTR CLI.

Usage Instructions

We provide example data in the example directory to demonstrate the usage of SSTR. You can use the example data to test the CLI commands.

De Novo Generation of Chemical Structures

To start the de novo generation, you must provide an MS/MS spectrum in either an MSP or MGF file format. The file must have a .msp or .mgf extension. One file can only contain one spectrum.

The essential properties required in the file are:

• FORMULA: Molecular formula of the compound.
• IONMODE: Ionization mode (positive or negative).
• PRECURSOR_MZ: Precursor mass/charge ratio.
• ADDUCT: The adduct form.

One example of such a file is provided in the example/lipid.mgf file.

To annotate the molecular formula, you can use external tools like Buddy or SIRIUS.

To generate one structure, run the following command:

sstr generate <path_to_msp_or_mgf_file>

You can enable stream mode to actually see the generation process:

sstr generate --stream <path_to_msp_or_mgf_file>

To generate 10 structures using beam search, run the following command:

sstr propose --beam 10 <path_to_msp_or_mgf_file>

Ranking Candidate Structures

To rank candidate chemical structures, provide the MS/MS spectrum file along with a file containing candidate structures in SMILES format.

A recommended approach is to annotate the molecular formula first, then retrieve the candidate structures with the same molecular formula from a database like PubChem.

The candidate SMILES should be stored in a .txt file, with one SMILES string per line.

One example of the candidate SMILES file is provided in the example/isomers.txt file.

To rank the candidate structures based on the MS/MS spectrum:

sstr rank <path_to_msp_or_mgf_file> --candidates <path_to_candidate_smiles_file>