Protein Binding

Overview

This module can be used to predict plasma protein bound fraction and the equilibrium binding constant to serum albumin of a compound. The binding properties are computed from physicochemical properties (lipophilicity, ionization constants, hydrogen bonding capacity) and structural descriptors of the compounds. Reliability of each prediction is assessed by calculating corresponding Reliability Indices (see Overview of the Trainable models).

Features

• Calculates %PPB – the cumulative percentage of the analyzed compound bound to human plasma proteins (such as albumin, alpha₁-acid glycoprotein and others) and log K_a^HSA – human serum albumin affinity constants. with Reliability Index (RI) values of the corresponding predictions.
• RI values represent a quantitative evaluation of prediction confidence. High RI shows that the calculated value is likely to be accurate, while low RI indicates that no similar compounds with consistent data are present in the training set.
• Main plasma proteins that mostly contribute to binding of different compound classes in human plasma are enumerated in the textual comments next to the prediction results.
• Experimentally measured extent of plasma protein binding, human serum albumin affinity constants are displayed for up to 5 similar structures from the training set along with the corresponding literature references.
• A clear and straightforward interface ensures quick and easy addition of user-defined data to the Self-training Library.

Interface

1. Calculated plasma protein bound fraction and logarithm of the affinity constant to human serum albumin in the blood plasma:
   File:distribution ppb scale.png
2. Reliability indices of predictions are provided along with the calculated values
3. A brief general description of a compound’s likely binding behavior when dissolved in blood plasma, based on its physicochemical properties
4. Up to 5 similar structures in the protein binding training set with experimental quantitative values of %PPB, LogK_a^HSA and literature references

Note: Prediction reliability classification according to Reliability Index (RI) values:

• RI < 0.3 – Not Reliable,
• RI in range 0.3-0.5 – Bordeline Reliability,
• RI in range 0.5-0.75 – Moderate Reliability,
• RI >= 0.75 – High Reliability

Technical information

Calculated quantitative parameters

Parameters calculated by Distribution\Protein Binding module include percentage plasma protein binding values (%PPB) and log K_a^HSA constants. These properties are related, but characterize provide slightly different information about the considered process.

• log K_a^HSA represents the drug’s affinity constant to human serum albumin – the major carrier protein in plasma. Experimental data come from direct chromatographic determination of binding strength to that particular protein.

log K_a^HSA = log ([LA]/([L][A])) where [LA] is concentration of ligand bound to albumin, [L] – that of free ligand, and [A] – concentration of free albumin which is estimated at ~0.6 mM in human plasma.

• %PPB values represent the overall fraction of drug bound in human plasma, i.e. accounts for interactions with different proteins: albumin, alpha₁-acid glycoprotein, lipoproteins, SHBG, transcortin etc. In vitro measurements of the extent of plasma protein binding usually involve equilibrium dialysis, ultrafiltration or ultracentrifugation methods.

%PPB = (1 – f_u) * 100% where f_u is fraction of free (unbound) drug in plasma ranging from 0 to 1.

• Supplementary Distribution\Vd module calculates apparent Volume of Distribution of drugs in human body expressed in liters per kg body weight (L/kg).

Experimental data

Experimental data that were utilized to build predictive models were collected from drug prescription information, reference pharmacokinetic tabulations and many original articles. The main sources of Vd data were well-known pharmacokinetic books: "Therapeutic Drugs" (ed. by C. dollery), and Goodman & Gilman's "The Pharmacological Basis of Therapeutics", while albumin affinity constants were collected mainly from original articles by Valko K. et al. J Pharm Sci. 2003;92(11):2236-48. [1], and Kratochwil N.A. et al. Biochem Pharmacol. 2002;64(9):1355-74. [2] The compiled data sets contain %PPB data for almost 1500 compounds, about 340 albumin affinity constants and almost 800 Vd values.

Model development (technical details)

The models for predicting %PPB and log K_a^HSA constants were developed with Algorithm Builder using a novel methodology consisting of two parts:

• Global baseline statistical model employing binomial PLS with multiple bootstrapping using a predefined set of fragmental descriptors.
• Local correction to baseline prediction based on analysis of experimental data for similar compounds.

The underlying methodology enables obtaining an intrinsic evaluation of prediction confidence by the means of Reliability Index (RI) values calculated for each prediction. RI ranging from 0 to 1 serves as an indication whether a submitted compound falls within the Model Applicability Domain. Two criteria influence the calculation od Reliability Index of a prediction:

• Similarity of the analyzed molecule to compounds in the Self-training Library (prediction is unreliable if no similar compounds have been found in the Library).
• Consistency of experimental data for similar compounds – reliability of calculated values is lower is data for similar compounds are discrepant.

Both %PPB and log K_a^HSA predictive models are Trainable meaning that their Applicability Domains may be easily extended by addition of ‘in-house’ experimental data to the module Self-training Library. Notably, the baseline statistical model does not need to be rebuilt from scratch to account for data entered by the user. The model is retrained automatically as new compounds are added to the Library. Model trainability would be particularly useful for predicting serum albumin affinity constants as literature data sets are very sparse, thus the ability to take advantage of large ‘in-house’ libraries gives the potential for a significant improvement of both accuracy and reliability of calculations.