Analysis of Algorithms Predicting Blood:Air and Tissue:Blood Partition Coefficients from Solvent Partition Coefficients for Prevalent Components of JP-8 Jet Fuel

Abstract

Algorithms predicting tissue and blood partition coefficients (PCs) from solvent properties were compared to assess their usefulness in a petroleum mixture physiologically based pharmacokinetic/pharmacodynamic model. Measured blood:air and tissue:blood PCs for rat and human tissues were sought from literature resources for 14 prevalent jet fuel (JP-8) components. Average experimental PCs were compared with predicted PCs calculated using algorithms from 9 published sources. Algorithms chosen used solvent PCs (octanol:water, saline or water:air, oil:air coefficients) due to the relative accessibility of these parameters. Tissue:blood PCs were calculated from ratios of predicted tissue:air and experimental blood:air values (PCEB). Of the 231 calculated values, 27% performed within +/- 20% of the experimental PC values. Physiologically based equations (based on water and lipid components of a tissue type) did not perform as well as empirical equations (derived from linear regression of experimental PC data) and hybrid equations (physiological parameters and empirical factors combined) for the jet fuel components. The major limitation encountered in this analysis was the lack of experimental data for the selected JP-8 constituents. PCEB values were compared with tissue:blood PCs calculated from ratios of predicted tissue:air and predicted blood:air values (PCPB). Overall, 68% of PCEB values had smaller absolute % errors than PCPB values. If calculated PC values must be used in models, a comparison of experimental and predicted PCs for chemically similar compounds would estimate the expected error level in calculated values.

Modeling of Complex Mixtures: JP-8 Toxicokinetics

“…For more minor components, it will likely prove prohibitive to measure each of these values. Therefore mathematical algorithms predicting tissue and blood partition coefficients (PCs) from solvent properties were compared to assess their usefulness in a petroleum mixture PBPK model (Sterner et al, 2006).…”

“…Overall, 68 percent of PCEB values had smaller absolute percent errors than PCPB values. If calculated PC values must be used in models, a comparison of experimental and predicted PCs for chemically similar compounds would estimate the expected error level in calculated values (Sterner et al, 2006).…”

Novel Models for Accurate Estimation of Air–Blood Partitioning: Applications to Individual Compounds and Complex Mixtures of Neutral Organic Compounds

“…Hence, regulatory bodies recommend the development of prediction models as viable alternatives to avoid extensive experimentation. , These models include Poly Parameter Linear Free Energy relationships (PP-LFERs) ,− Quantitative Structure–Activity Relationship (QSAR) Models, , Artificial Neural Network (ANN) Models, Random Forest (RF) Models, Genetic Algorithm (GA) Models, and hybrid approaches which are combination of two or more methods . However, each model has inherent limitations.…”

“…28 , 29 These models include Poly Parameter Linear Free Energy relationships (PP-LFERs) 2 , 30 − 32 Quantitative Structure–Activity Relationship (QSAR) Models, 33 , 34 Artificial Neural Network (ANN) Models, 35 Random Forest (RF) Models, 36 Genetic Algorithm (GA) Models, 35 and hybrid approaches which are combination of two or more methods. 37 However, each model has inherent limitations. PP-LFERs may necessitate extensive experimental data for parametrization and may lack accuracy in predicting complex chemical interactions.…”