Modeling Cancer Cells Growth

Abstract

In molecular oncology the analysis of growth patterns of cells represent a major goal for understanding the evolution of growth profiles in pathological conditions. To compare various experimental settings in a pre-clinical study of inhibitors of prostate cancer, we modeled cancer cell growth using different statistical approaches which account for biological variability and complexity within experiments and over time. We first estimated cell growth by means of Linear Mixed Effects models that included unobserved factors that may influence cell development. Random effects were included to allow for no constant variance and covariance of residuals over time (Wu and West, 2009) and to properly model the whole growth process (i.e., all observation times). These are crucial aspects that are commonly ignored by the standard Linear Model. Since the nature of the data does not support the assumption of strictly linear effect of the continuous covariates on the predictor we present a solution within generalized additive models with mixed effects. An extension of this modeling to a fully Bayesian framework is also considered as it has a high degree of flexibility.

Keywords:

• Bayesian approach
• Generalized additive models
• Growth curve modeling
• Linear mixed effects model

Mathematics Subject Classification:

• Primary 62F15
• Secondary 62P10

View correction statement:

Erratum

Acknowledgments

The authors wish to thank the Editor and two anonymous referees for their thorough and constructive review.

Notes

siRNA techniques were used to investigate the effect of Myo6 genetic depletion on the phenotype and tumorigenicity of androgen-independent PC3 cells (Helenius et al., 2001).

gamm function in R mgcv package, returns a lme object (for fixed and mixed effects) and a gam object (for the smooth terms).