chemlog 1.0.6

Peptide classifier for ChEBI / PubChem

ChemLog is a framework for rule-based ontology extension. This repository implements a classification of peptides on the ChEBI and PubChem datasets.

Installation

You can install ChemLog with pip:

pip install chemlog

To get the latest development version, download the source code and install with

pip install .

If you want to use the MONA reasoner, you have to install it separately (the classifier expects the mona command to be available).

Run the classification

ChemLog provides a command line interface for the classification. Results are in JSON format for each run, alongside a log and a config file. Currently, classification of ChEBI and PubChem data is supported. Download and preprocessing of the data are handled automatically. For instances, the following command classifies the 1,000 smallest peptides in ChEBI with the algorithmic method:

python -m chemlog classify-chebi --chebi-version 239 --strategy algo --only-peptides --n-molecules 1000

For more details on the available command line options run

python -m chemlog --help

Publication

Flügel et al. (2025): ChemLog: Making MSOL Viable for Ontological Classification and Learning

How are peptides classified?

4 methods for classification are implemented:

1. Using Monadic Second-Order Logic (MSOL) formulas with the MSOL model finder MONA
2. Turning an MSOL model finding problem into a QBF satisfiability problem and solving that with CAQE or DepQBF, using the Bloqqer preprocessor.
3. Turning an MOSL model finding problem partially into First-Order Logic (FOL) and solving that with a custom FOL model checker (since not all MSOL axioms are translatable, the non-translatable parts are calculated algorithmically).
4. Using an algorithmic implementation

If you are just interested in the results, we recommend choosing the algorithmic implementation, as it is the fastest and can handle complex molecules.

The classification covers the following aspects:

1. Number of amino acids (up to 10, except for the algorithmic method, which covers arbitrary sizes)
2. Charge category (either salt, anion, cation, zwitterion or neutral)
3. Proteinogenic amino acids present
4. Emericellamides and 2,5-diketopiperazines

ChemLog will also return the ChEBI classes that match this classification. Currently supported are:

ChEBI ID	name
16670	peptide
60194	peptide cation
60334	peptide anion
60466	peptide zwitterion
25676	oligopeptide
46761	dipeptide
47923	tripeptide
48030	tetrapeptide
48545	pentapeptide
15841	polypeptide
90799	dipeptide zwitterion
155837	tripeptide zwitterion
64372	emericellamide
65061	2,5-diketopiperazines
24866	salt
25696	organic anion
25697	organic cation
27369	zwitterion

All implementations are based on the same natural language definitions and have been developed jointly. Therefore, it is expected that all methods yield the same result. If you make a different experience, please open an issue.

If you face problems using ChemLog or have other questions, feel free to open an issue as well.