Digitized dataset of aqueous acid dissociation constants

Area of Science:

• * Chemistry
• * Chemical Informatics

Background:

• * The acid dissociation constant (pKa) is vital for drug design, environmental studies, and synthesis.
• * Limited availability of high-quality, open-source digital pKa datasets hinders research and predictive modeling.
• * Existing datasets often suffer from data overlap issues, compromising the reliability of pKa predictors.

Purpose of the Study:

• * To release the IUPAC Digitized pKa Dataset, a comprehensive, FAIR open-source resource.
• * To detail the digitization and validation process of historical pKa data.
• * To develop and evaluate a pKa predictor using the new dataset with overlap-free testing.

Main Methods:

• * Digitization and critical assessment of historical pKa data compiled up to 1970.
• * Inclusion of metadata: temperature, measurement method, reliability, and chemical identifiers (SMILES, InChI).
• * Training a macroscopic pKa predictor and validating its accuracy on an independent, overlap-free test set.

Main Results:

• * The release of the largest FAIR open-source aqueous pKa dataset, containing 24,222 entries for 10,564 unique molecules.
• * Comprehensive data checking and assessment of the dataset's informational space.
• * Development of a pKa predictor with validated accuracy on unseen data, addressing issues of training-test overlap.

Conclusions:

• * The IUPAC Digitized pKa Dataset significantly enhances the availability of reliable pKa data for scientific research.
• * This resource facilitates advancements in data-driven pKa prediction and related chemical applications.
• * The study provides a robust methodology for developing and validating predictive models using curated datasets.