Identifying uncertainty in physical–chemical property estimation with IFSQSAR

Trevor N. Brown,Alessandro Sangion,Jon A. Arnot

DOI: https://doi.org/10.1186/s13321-024-00853-w

2024-06-02

Journal of Cheminformatics

Abstract:This study describes the development and evaluation of six new models for predicting physical–chemical (PC) properties that are highly relevant for chemical hazard, exposure, and risk estimation: solubility (in water S W and octanol S O ), vapor pressure ( VP ), and the octanol–water ( K OW ), octanol–air ( K OA ), and air–water ( K AW ) partition ratios. The models are implemented in the Iterative Fragment Selection Quantitative Structure–Activity Relationship (IFSQSAR) python package, Version 1.1.0. These models are implemented as Poly-Parameter Linear Free Energy Relationship (PPLFER) equations which combine experimentally calibrated system parameters and solute descriptors predicted with QSPRs. Two other ancillary models have been developed and implemented, a QSPR for Molar Volume ( MV ) and a classifier for the physical state of chemicals at room temperature. The IFSQSAR methods for characterizing applicability domain (AD) and calculating uncertainty estimates expressed as 95% prediction intervals (PI) for predicted properties are described and tested on 9,000 measured partition ratios and 4,000 VP and S W values. The measured data are external to IFSQSAR training and validation datasets and are used to assess the predictivity of the models for "novel chemicals" in an unbiased manner. The 95% PI intervals calculated from validation datasets for partition ratios needed to be scaled by a factor of 1.25 to capture 95% of the external data. Predictions for VP and S W are more uncertain, primarily due to the challenges in differentiating their physical state (i.e., liquids or solids) at room temperature. The prediction accuracy of the models for log K OW , log K AW and log K OA of novel, data-poor chemicals is estimated to be in the range of 0.7 to 1.4 root mean squared error of prediction (RMSEP), with RMSEP in the range 1.7–1.8 for log VP and log S W .

chemistry, multidisciplinary,computer science, interdisciplinary applications, information systems

What problem does this paper attempt to address?

The paper attempts to address the issue of identifying and quantifying uncertainty in the prediction of physicochemical properties of chemical substances. Specifically, the study developed 6 new models to predict physicochemical (PC) properties crucial for chemical hazard, exposure, and risk assessment: water solubility (SW), octanol solubility (SO), vapor pressure (VP), and octanol-water (KOW), octanol-air (KOA), and air-water (KAW) partition coefficients. These models are achieved by integrating experimentally calibrated system parameters and solute descriptors predicted using quantitative structure-activity relationships (QSPR). Additionally, the study developed 2 auxiliary models: one for predicting molar volume (MV) and another for classifying the physical state of chemicals at room temperature (liquid or solid). The focus of the research is on evaluating the models' predictive capabilities for "novel chemicals" using external datasets and testing the models' predictive power and their uncertainties in practical applications. The results indicate that for novel, data-poor chemicals, the prediction error range for octanol-water, air-water, and octanol-air partition coefficients is estimated to be between 0.7 and 1.4; whereas for vapor pressure and water solubility, the prediction error range is between 1.7 and 1.8. This suggests that there remains significant uncertainty in the prediction of certain physicochemical properties.

Identifying uncertainty in physical–chemical property estimation with IFSQSAR

Improved prediction of PFAS partitioning with PPLFERs and QSPRs

QSPR models for the physicochemical properties of halogenated methyl-phenyl ethers.

Developing the QSPR Model for Predicting the Storage Lipid/water Distribution Coefficient of Organic Compounds

QSPR models to predict the physical hazards of mixtures: a state of art

QSPR Study of N‐octanol/water Partition Coefficient of Some Aromatic Compounds Using Support Vector Regression

Selecting reliable physicochemical properties of perfluoroalkyl and polyfluoroalkyl substances (PFASs) based on molecular descriptors

The system of self-consistent models for pesticide toxicity to Daphnia magna

Gas-to-ionic Liquid Partition: QSPR Modeling and Mechanistic Interpretation.

Development and Evaluation of Conformal Prediction Methods for QSAR

How Precise Are Our Quantitative Structure-Activity Relationship Derived Predictions for New Query Chemicals?

QSPR analysis of phenylthio phenylsulfinyl and phenylsulfonyl esters using quantum chemical semi-empirical descriptors

QSPR models for n-octanol/water partition coefficient and enthalpy of vaporization using CDFT and information theory-based descriptors

Chemical Space Covered by Applicability Domains of Quantitative Structure-Property Relationships and Semiempirical Relationships in Chemical Assessments

QSPR models for solvation enthalpy based on quantum chemical descriptors

p Ka prediction of per- and polyfluoroalkyl acids in water using in silico gas phase stretching vibrational frequencies and infrared intensities

Novel Bayesian Method to Derive Final Adjusted Values of Physicochemical Properties: Application to 74 Compounds

Methods for comparing uncertainty quantifications for material property predictions

Property estimation of organic compounds based on QSPR models with norm indices

Prediction of the hydrophobicity of platinum(IV) complexes based on molecular surface properties

Prediction of n-octanol/water partition coefficients and soil sorption coefficients for polychlorinated organic compounds (PCOCs) using quantum chemical parameters