High Throughput Screening For Drug Distribution In Vivo

Recent studies have shown that about 10%promising new drug candidates were abandoned due to poor pharmacokinetic properties（PKs）.Therefore,it is desirable to develop effective and accurate methods for estimating the pharmacokinetic properties of a candidate drug as early as possible during drug discovery and development.One important aspect of the pharmacokinetic properties（PKs） of a drug candidate is its distribution performance. Reasonally understanding of distribution behavior contributes to predict drug’s pharmacological action,the extent of residense in vivo,and side effect,to ensure safe medication,and to promote the development of new drugs.The distribution performance are usually described by the volume of distribution（Vd） and tissue:plasma partition coefficients （Kp）.Vd at steady state(Vd_ss) is an important parameter to calculate the mean residence time and half-life of the drug,indicative of the drug exposure profile and the dosing frequency for setting up a suitable dosing regimen.The methods used to predict Vd include（ⅰ） the extrapolation of animal data,（ⅱ） physiologically based pharmacokinetic（PBPK） modeling and（ⅲ） in silico approaches that employ quantitative structure - pharmacokinetic relationships.Kp refelct drug’s tissue distribution or accumulation trend in vivo.It is an important model parameter for PBPK,rapid and accurate determintation of Kp will promote PBPK model’s implemention in the drug development phase.The predicted approaches were studied mostly on mechanistic equations based on tissue composition.Considered that compound’s distribution in vivo mainly depended on it’s physicochemical properties,and physicochemical parameters had clarifies meaning and were acquiring easily,we used them as the mainly parameters to construct three kinds of predicted Vd_ss models and one kind of predicted Kp model.The studied contents are as follows:Predicted Vd_ss modelâ… - Extrapolation of Vd_ss in human from the preclinical animal data in combinated with physicochemical preproties（short for animals’ data method）Dataset included the Vd_ss values for 121 drugs in rat,dog,monkey and human,collected from literatures,and seven physicochemical properties calculated by ACD/labs soflward. Several linear predictive models were constructed by partial least squares（PLS） using three or one animal data in combination with or without the physicochemical descriptors.Stepwise regression was performed to choose the more important physicochemical descriptors.The predictive accuracy of PLS model with three animals data was better than the allometric method（predicted AFE was 1.91 and 4.75,respectively）.Moreover,the predictive accuracy of these models were improved when the physicochemical descriptors were introduced. Interestingly,the common contributors selected were molecule weight（MW） and the negatively charged fraction value(f_i-),which were considered as the significant parameters and the latter was considered in the extrapolation process for the first time.The model with three animals data,MW and f_i- had the most accurate prdicted result.Exterior validation was performed and verified the powful predictablity of this model.Compared to the similar model reported,our model has greater predicted accuracy.In order to validate our Vd_ss model’s stability and predictability,the validation set was consisted of the 14 patent compounds.After 14 patent compounds were injected into rat tail vein,respectively,the compound’s concentration in plasma samples were determined by HPLC-MS analysis.Statistical moment analysis were performed to calculte Vd_ss values.In addition,the equilibrium dialysis were carried out to plasma protein binding rate（f_u） of these 14 compounds in rat plasma.The Vd_ss and f_u of 14 compound were reserved.Predicted Vd_ss modelâ…¡- Prediction of Vd_ss using unbound drug fraction in tissues(f_ut) (short for logk_IAM method)By multiple regression analysis with log k_IAM（logarithmic retention factor in immobilized artificial membrane chromatography）,together with f_i（7.4） and log f_u,the predictive equation of f_ut（the fraction of the compound unbound in tissues） for the 121 compounds in humans and 82 compounds in rats were built,predictive Vd_ss was further obtained from the（?）ie-Tozer equation.Exterior validation（test set and validation set） and interior validation （leave-class-out） were performed.The results indicated the model had a robust predictive ability(R² for f_ut equation＞0.8,AFE≤2 for Vd_ss in test set and in validation set).Compared to the literatures’ methods,our model had better predictive accuracy and wider application.Predicted model for Kp using physicochemical parametersAccording to compounds protolytic properties,116 compounds data were divided into two subsets,one for acids and zwitterions and another for bases and neutrals.Each of these two datasets were then subdivided randomly into one training set and one test set（4:1） according cluster analysis reslut.In training set of 91 compounds,correlations between Kp_muscle and physicochemical parameters were obtained by stepwise regression,correlations between Kp values for muscle and other 12 tissues were improved when fi was introduced into.Then predicted model for Kp using physicochemical parameter were constructed.For a test set of 25 compounds,predicted Kp_muscle values and Kp_tissue agreed well with experimental values （AFE＜1.7 and AFE≈2,respectively）.Compared to the literatures’ methods,our model had similar or better predictive accuracy,and could be easilier and convenientlier applied. Predicted Vd_ss modelâ…¢- Predicted model for Vd_ss using physicochemical parametersTo use developed physicochemical model to predict Kp_muscle in rat,apply these predictions to determinate Vd_ss in rat.Kp_muscle in rat assumed was similar to Kp_muscle in human,Vd_ss in human was determinated.Vd_ss values were predicted for 92 structurally diverse compounds in rats and 82 in humans.Exterior validation(14 patent compounds’ Vd_ss in rat) was performed to verify the method’s steady and predictability.The prediction accuracy was good in rat and human.AFE was less than 2 for acids and zwitterions,AFE was slightly larger than 2 for bases and neutrals,prediction accuracy for the former was greater than the latter.For validation set,AFE was 1.87.This approach only used compounds’ physicochemical parameter to predict Vd_ss value,was simple and rapid.Finally,predictablity and applicatable phase of three kinds of Vd_ss predicted methods were compared and summaried.Animals’ data method（Ⅰ） and logk_IAM method（Ⅱ） have greater predicted accuracy than predicted mothods using physicochemical parameters（Ⅲ）,the two later have high throught screening property,and predicted speed of the later was rapider. Three models were suitable to apply in different stage of new drug research and development, to guide to design clinical plan in the preclinical phase,to high throught screen candidated drug’s pharmacokinetics properties in early phase,and to screen candidated drug and guide to synthesis perfect candidates in the initial stage.