HOME     SCHEDULE     PROGRAM     AUTHOR INDEX     SUBJECT INDEX               SIGN UP

PARENT SESSION

Risks and Modeling at Low Doses

Sunday, October 16, 2005 3:00 PM-5:00 PM Exhibit Hall

(PP051) Fold change does not accurately assess transcriptional change due to radiation exposure.

Rocke, David*,1, Goldberg, Zelanna2, Schweitert, Chad2, Santana, Alison2, Jones, Angela2, Stern, Robin, Lehmann, Joerg2, 1 Division of Biostatistics, Davis, CA, USA2 Department of Radiation Oncology, Davis, CA, USA

ABSTRACT- The most commonly used method of assessing transcriptional change in radiation biology, as in most areas of biology, is fold change, which is the ratio of the measured response in one condition to that in another. Sometimes this is also used in the form of a log ratio. The most serious flaw in the use of this summary method is that there is no yardstick by which one can distinguish a large fold change from a small one. Biologists may assume that a fold change of 2 is validated by some statistical principle, and statisticians may assume that this criterion is validated by some biological principle, but in fact neither is true, and the rule has no scientific basis whatsoever. What is missing is some measure of what variation in transcript level would occur without the radiation exposure. This clearly differs from transcript to transcript, so a gene-specific comparison is needed, such as that provided by standard statistical methods such as the t-test, the analysis of variance, or regression analysis. Other problems with fold change include 1) it can only conveniently be used when evaluating a single factor at two levels, and increasingly tortuous adaptations are needed for more complex experimental designs; 2) it fails to make sense in the important case in which transcription is essentially absent in one experimental condition, since that implies a zero denominator; 3) it results in variance inhomogeneities that make application of standard statistical methods less effective or even causes them to give fallacious results. We illustrate the superior effectiveness of alternative analysis methods based on solid statistical principles using self-self hybridizations, cell-line radiation exposure data, and controlled human low-level radiation exposure data.

Key words: microarray, gene expression, fold change, mRNA


Internet Services provided by
Allen Press, Inc. | 810 E. 10th St. | Lawrence, Kansas 66044 USA
e-mail assystant-helpdesk@allenpress.com | Web www.allenpress.com
2005 RRS