How to crack a SMILES: automatic crosschecked chemical structure resolution across multiple services using MoleculeResolver

Abstract: Accurate chemical structure resolution from textual identifiers such as names and CAS RN® is critical for computational modeling in chemistry and related fields. This paper introduces MoleculeResolver, an automated, robust Python-based tool designed to address inconsistencies and inaccuracies commonly encountered when converting chemical identifiers to canonical SMILES strings. MoleculeResolver systematically crosschecks structures retrieved from multiple reputable chemical databases, implements rigorous identifier plausibility checks, standardizes molecular structures, and intelligently selects the most accurate representation based on a unique resolution algorithm. Scientific contribution: Benchmarks across diverse datasets confirm that MoleculeResolver significantly enhances precision, recall, and overall reliability compared to traditional single-source methods, proving its utility as a valuable resource for chemists, data scientists, and researchers engaged in high-quality molecular data analysis and predictive model development.

Subjects

Chemical structure retrieval

Identifier

ML

MoleculeResolver

Python

QSPR

SMILES

DDC Class

540: Chemistry

620: Engineering

Publication version

publishedVersion

Name

s13321-025-01064-7.pdf

Type

Main Article

Size

3.58 MB

Format

Adobe PDF

Options

How to crack a SMILES: automatic crosschecked chemical structure resolution across multiple services using MoleculeResolver