Statistical models for genetic data in relation to cancer research

In cancer research, there has been a long time interest in identifying genes or allelic markers differentially expressed in cases and controls. By attaching a fluorescent colour to the primer of interest, it is possible to monitor the increase in PCR product of the cases versus controls using real time quantitative PCR (qPCR). A significant difference in the hitting times (in PCR cycles) between cases and controls is indications of an associations between the disease and genetic marker.

The currently applied statistical test for evaluating this difference while taking the uncertainty into account is based on a normal approximation of the hitting time. However, using the theory from stochastic processes it is not clear that this assumption is valid. Preliminary experiments suggest that the tails of the distribution is heavier than those of a normal distribution leading to erroneously narrow confidence intervals for the difference. As a consequence will the probability of a false positive be higher than the suggested significance level.