Da. Kramer et Rb. Conolly, COMPUTER-SIMULATION OF CLONAL GROWTH CANCER MODELS .1. PARAMETER-ESTIMATION USING AN ITERATIVE ABSOLUTE BISECTION ALGORITHM, Risk analysis, 17(1), 1997, pp. 115-126
Quantitative models of the relationship between exposure to chemical c
arcinogens and carcinogenic response are useful for hypothesis evaluat
ion and risk assessment. The degree to which such models accurately de
pict the underlying biology is often a function of the need for mathem
atical tractability. When closed-form expressions are used, the need f
or tractability may significantly limit their complexity. This problem
can be minimized by using numerical computer simulation methods to so
lve the model, thereby allowing more complex and realistic description
s of the biology to be used. Unfortunately, formal methods of paramete
r estimation for numerical models are not as well developed as they ar
e for analytical models. In this report, we develop a formal parameter
estimation routine and apply it to a numerical clonal growth simulati
on (CGS) model of the growth of preneoplastic lesions consisting of in
itiated cells. An iterative bisection algorithm was used to estimate p
arameters from time-course data on the number of initiated cells and t
he number of clones of these cells. The algorithm successfully estimat
ed parameter values to give a best fit to the observed dataset and was
robust vis-a-vis starting values of the parameters. Furthermore, the
number of data points to which the model was fit, the number of stocha
stic repetitions and other variables were examined with respect to the
ir effects on the parameter estimates. This algorithm facilitates the
application of CGS models for hypothesis evaluation and risk assessmen
t by ensuring uniformity and reproducibility of parameter estimates.