JGuiB - AN ASSISTANT TOOL FOR PHARMACOKIN ETIC/PHARMACODYNAMIC MODELING WORKING WITH BOOMER Ying-hao Lu 1 *, David W.A. Bourne 2 , Yung-jin Lee 1 1 College of Pharmacy, Kaohsiung Medical University, Kaohsiung, Taiwan 2 College of Pharmacy, University of Oklahoma Health Sciences Ce nter (OUHSC), Oklahoma City, OK 73119, U.S.A. Introduction BOOMER (www.BOOMER.org ) created by Dav id W.A. Bourne is an excellent non-lin ear regression tool for data analysis and modeling of pharmacokinetic (PK) a nd pharmacodynamic (PD) data. 1 It is v ery suitable for purposes of computer- aided PK/PD modeling teaching and data analysis with many built-in numerical integration and nonlinear optimization algorithms. Unfortunately, BOOMER onl y runs in a menu driven mode under dif ferent platforms. Therefore it may ta ke time for a beginner to learn how to use it in PK/PD modeling. The purpose of this study was to build a GUI (Grap hical User Interface) for BOOMER to as sist users who need a GUI to manipulat e the operation of BOOMER. Keywords: BOOMER, JGuiB, PK/PD Modeling, Non-linear Regression, Java, Java SE Methods and Materials When starting to build the GUI, the fir st problem we had to solve was to modif y BOOMER to accept command-line instruc tions. In order to make the GUI work w ith BOOMER under background, BOOMER was modified to be able to run under comman d-line mode. This allowed the operatio n of BOOMER controlled by the GUI gener ated batch files. Since BOOMER has the cross-platform property to run on Mac O S X, Windows and Linux, our GUI should exploit this property. Thus we selecte d Java (Java SE) as programming languag e to build our GUI, and used JBuilder X Foundation, a powerful Java IDE (Integr ated Development Environment), as devel oper and compiler. We call this GUI as JGuiB, which stands for Java GUI for BO OMER. Results and Discussion We have successfully turned the menu-d riven mode of BOOMER into a GUI-based application with JGuiB. JGuiB with BO OMER can run in platforms of Mac OS X, Windows and Linux. Three main BOOMER functions in PK/PD modeling are suppor ted: normal fitting, simulation and Ba yesian estimation. Users just follow the sequences of PK/PD modeling (model building/selection), data manipulation (spreadsheet-like data editing; copy a nd paste) and weighting scheme selecti ons, and finally numerical integration /nonlinear optimization selections thr ough GUI and run this project with BOO MER. Twenty four frequently used PK/P D models with graphics were built and included with JGuiB. JGuiB also allow s users to build their defined PK/PD m odels with a panel of“Parameter Table”, such as physiologically based pharmac okinetic (PBPK) model. To avoid possi ble input errors made by users, JGuiB is designed to lessen user’s inputs, s uch as without assigning “lower limit” and “upper limit” for each adjustable parameter. Two of more fascinating fu nctions of JGuiB are that: (1) it prov ides the clinical-oriented Bayesian es timation functions with multiple-dosin g modeling functions for uniform or no n-uniform doses, 2 and (2) it enables BO OMER to simultaneously work with diffe rent PK/PD models (max. 4 models in on e project file) using the same set of data for the purpose of model discrimi nation (Figure 1). We contrast BOOMER with JGuiB plus BOOMER in Table 1, 3 and compare JGuiB plus BOOMER with WinNonl in in Table 2. 4 Finally there are the screenshots of JGuiB to show how to us e it with friendly-to-use functions in Figure 2. Conclusion and Future Wor ks JGuiB has been successfully built and it is just a GUI, an assistant tool f or BOOMER in PK/PD modeling. It gene rates a batch file (Input file) which is subsequently analyzed using BOOMER which provides the computational engi ne. Therefore, JGuiB is dependent on BOOMER. The built-in PK model librar y in JGuiB enables users to apply the se models quickly to analyze data wit h BOOMER. Boomer also can easily app ly to multiple dosing modeling and Ba yesian estimation into the field of c linical pharmacokinetic services with excellent model building templates an d easy-to-use data management of JGui B. Therefore, JGuiB effectively exte nds the abilities of Boomer. In conc lusion, JGuiB is the first GUI progra m for individual PK/PD software in th e world. In the future, we plan to e nhance the modeling automation of JGu iB to include PBPK modeling and Monte Carlo simulation. In addition, we wi References 1.Bourne DWA, BOOMER, 1989, a simulation and modeling program for pharmacokinetic and pharmacodynamic data analysis. Computer Methods and Programs in Biomedicine, 29: 191-95 2.Yamaoka K, Nakagawa T, Tanaka H, Yasuhara M, Okumura K, Hori R, 1985, A nonlinear multiple regression program, MULTI2 (BAYES), based on Bayesian algorithm for microcomputers. Journal of Pharmacobio-Dynamic, 8: 246-56. Table 1 - Comparison of BOOMER and JGuiB (+ BOOMER) 3 Program Format Coding Language Platform # MacOS MacOS X Windows Linux BOOMER Menu-Driven FORTRAN Yes Yes Yes Yes JGuiB (+ BOOMER) Graphic Interface JAVA No Yes Yes Yes Program Model No. Subject No. Data Management Error Check Lower Limit/Upper Limit Help Function BOOMER 1 * 1 * From Keyboard a Few Users supply Online Manual (www.BOOME R.org) From Data file JGuiB (+ BOOMER) max. 4 ** no limitation ** Excel-Formatted Data Exchange Complete JGuiB Generates (10 - 8 /10 +8 ) Some Helpful functions *** copy & paste Program PK/PD Modeling Functions Normal Fitting Normal Simulation Bayesian Estimation Monte-Carlo Simulation Multiple Dosage Regimens BOOMER Yes Yes Yes Yes Yes JGuiB (+ BOOMER) Yes Yes Yes No Yes * from each batch file ** from each project file (JGuiB can generate many batch files) *** a. support some project files for examples. b. visualized model graphs in built-in library. c. helpful video demo when linking to JGuiB website (http://jguib.pkpd.org.tw/). d. parameter type list s, parameter properties lists, error type lists. # BOOMER Version (Mac OS X, Windows, Linux: 3.3.2), JGuiB Version (2.1) Table 2 - Comparison of JGuiB (+ BOOMER) and WinNonlin 4 Program Version Reviewed Graphic Interface Platform Initial Parameter Value Lower Limit/ Upper Limit MacOS X Windows Linux JGuiB (+ BOOMER) 2.1 Yes Yes Yes Yes Users Supply JGuiB Generates WinNonlin 4.1 Yes No Yes No Users Supply WinNonlin Generates WinNonlin Generates Program Excel-compatible spreadsheet Plotting AUC/AUMC Calculation Print Charge JGuiB (+ BOOMER) Yes Yes Linear Trapezoidal Rule Yes $59.95 Method 9 of R.D. Purves WinNonlin Yes Yes Linear Trapezoidal Rule Yes $1000/ year Program Nonlinear Optimization Weighting Scheme* JGuiB (+ BOOMER) Gauss Newton Algorithm Simplex Algorithm Equal Weighting 1/Y 1/Y^2 Marquardt Algorithm Damping Gauss Newton Algorithm 1/a*Y^b 1/(a+b*Y^c) 1/[(a+b*Y^c)*d^(t n-ti)] Simplex to Damping Gauss Newton Algorithm WinNonlin Gauss Newton Algorithm Simplex Algorithm Equal Weighting 1/Y 1/Y^2 1/Y^(1/2) Program PK/PD Modeling Functions PK Model PD Model PK/PD Model PBPK ** Model NCA *** Model Macro-constant Exponential Model JGuiB (+ BOOMER) Yes Yes Yes Yes Yes Yes WinNonlin Yes Yes Yes No Yes Yes Program PK/PD Modeling Functions Built-in Library User’s Defined Library Bayesian Estimation Definition No. JGuiB (+ BOOMER) Differential Equation 24 Differential Equation Yes WinNonlin Integrated Equation 37 Differential Equation No * BOOMER allows explicit entry of weight as third data column, but JGuiB just supports the weighting scheme selections for the data sets. a, b, c, d - constant ** Physiologically-Based Pharmacokinetic *** Non-Compartmental Analysis Figure 1 - JGuiB (+ BOOMER) can generates 20 output files in the same subject data working with 4 PK/PD models. Figure 2 - the screenshots of JGuiB to show how to use it with friendly-to- use functions. 3.Boomer Website, 2006, http://www.BOOMER.org 4.WinNonlin Website, 2006, http://www.pharsight.com/products /winnonlin/