在R的環境下建立一套藥物動力學的模式-PKfit

Develop a Pharmacokinetic Program (PKfit) for R

李淳愔(Chun-Ying Lee)

指導教授 ： 李勇進

高雄醫學大學/藥學院/臨床藥學研究所/碩士(2006年)

摘要

本研究的目的為設計一個在R (www.r-project.org)的環境執行曲線藥物動力學非線性迴歸的程式。這個程式利用數個演算法，包括遺傳演算法來計算藥物動力學參數值，於研究中稱這個工具為PKfit。本研究還將所有從PKfit所得到的結果和其他兩個藥物動力學程式相比較：WinNonlin (www.pharsight.com)和Boomer (www.boomer.org)。
利用lsoda(來自odesolve套件)功能來處理定義藥物動力學模式的微分式。PKfit包含了三個不同的演算法: Gauss-Newton法用於非線性迴歸(nls函數)、Nelder-Mead法用於加權平方和的最小化(optim函數)，以及遺傳演算法(genoud函數來自rgenoud套件)。由於本研究的主要設計目的是要方便使用者的使用，因此發展選項式的介面，讓沒有任何程式經驗的使用者容易使用。PKfit提供了十四項藥物動力學模式：靜脈途徑給藥包括快速注射和靜脈輸注，血管外途徑給藥，一階次/零階次吸收，線性和非線性Michaelis-Menten排除藥物動力學模式。PKfit也提供兩種加權的方法: 1/Cp和1/Cp2。 輸出資料包括簡要的表格(包括時間、觀察和計算濃度、加權殘差、曲線下面積和第一動差曲線下面積)、最終的藥物動力學參數值、模式選擇條件值(Akaike's Information Criterion (AIC), Schwarz's Bayesian Criterion (SBC) 和 Log likelihood)和nls的診斷圖(包括線性圖、半對數圖以及殘差圖)。
本研究成功地在R上建構了PKfit，並將PKfit上傳至CRAN (http://cran.r-project.org)網站。PKfit，如同R，是一個授權於GNU Public License的開放原始碼套件。

關鍵字

藥物動力學 ； 非線性迴歸 ； 模型擬合 ； 模擬 ； 遺傳演算法

並列摘要

The purpose of this study was to create an R (www.r-project.org) program to fit nonlinear pharmacokinetics (PK) regression models using a variety of algorithms including a genetic algorithm. We call this tool PKfit. Results obtained from PKfit would be compared with those from other two available PK programs: WinNonlin (www.pharsight.com) and Boomer (www.boomer.org).
We used the lsoda function (from the odesolve package) to solve all differential equations used to define the PK models. PKfit includes three different data fitting algorithms: Gauss-Newton for non-linear regression (nls function); Nelder-Mead simplex for minimization of weighted sum of squares (optim function); and a genetic algorithm (genoud function from the rgenoud package). Since one of our key design goals was ease of use, we developed a menu-driven interface which does not require any programming experience to use. Fourteen pharmacokinetic models were implemented in PKfit: intravenous drug administrations with i.v. bolus or i.v. infusion, extravascular drug administrations, linear (1st-order absorption/elimination) and the nonlinear Michaelis-Menten model. Two weighting schemes, 1/Cp(obs) and 1/Cp2(obs), were also implemented. The output from PKfit includes a summary table (consisting of time, observed and calculated concentrations, weighted residuals, area under plasma concentration curve, and area under the first moment curve), final PK parameter values, model selection criteria (Akaike's Information Criterion (AIC), Schwarz's Bayesian Criterion (SBC) and Log likelihood) and diagnostic plots for nls such as linear plot, semi-log plot, and residual plot.
We have successfully built the PKfit package for R, and have uploaded it to the CRAN (http://cran.r-project.org) website. PKfit is an open source package licensed under the GNU Public License (http://www.gnu.org/coptleft/gpl.html) like R.

並列關鍵字

R ； Pharmacokinetics ； Nonlinear regression ； Model fitting ； Simulation ； PKfit

參考文獻

Akaike, H., A New Look at the Statistical Model Identification. IEEE Transactions on Automatic Control, 19:716-23, 1973.

Bates, D.M. and Watts, D.G., Nonlinear Regression: Iterative Estimation and Linear Approximates. Nonlinear Regression Analysis and Its Applications, John Wiley & Sons, Inc., USA, 32-66, 1988.

Becker, R.A., Chambers, J.M. and Wilks, A. R., The New S Language, Wadsworth & Brooks/Cole, 1988.

Best, N., Cowles, M.K. and Vines, K., CODA: Convergence Diagnosis and Output Analysis Software for Gibbs sampling output, Version 0.40.

Boroujerdi, M., General Introduction to Physiologic Modeling and Compartmental Analysis. Pharmacokinetics: Principles and Applications 1st Edition. McGraw-Gill Companies, Inc., New York, USA, 39-49, 2001.