DESIGN & ANALYSIS OF MICROARRAY GENE EXPRESSION STUDIES IN PLANTS: TOWARD SOUND STATISTICAL PROCEDURES

Microarrays are a new and exciting technology that offers plant biologists the opportunity to measure the relative amount of expression of literally thousands of genes simultaneously. This technology potentially allows plant biologists to make major leaps forward by increasing the number of factors they can examine and, importantly, offering the opportunity to study the complex relations and physiological coordination among entire genomes rather than single genes in isolation. It can address questions such as:

• At the genomic level, what is the cascade of adaptive or defensive events that a plant expresses when exposed to an environmental stress such as drought or pests?
• What is the relation between expression of specific genes and nutritionally important qualities, such as fatty acid composition, of soybeans?
• For which genes is enhanced or decreased expression related to superior yield? To name a few.

For this technology to achieve its potential, a richer understanding of the statistical properties of the data produced and statistically sound methods for analyzing those data must be developed and made available to plant biologists. Writings on analysis of microarray data have been rapidly passing through overlapping phases of increasing sophistication. Earliest writings consisted almost exclusively of simple descriptive statistics and the use and development of a variety of clustering routines without rigorous presentation of the statistical properties of the procedures. This was followed by a recognition that there is more than clustering to be done and inferential procedures that accommodate massive multiple testing are needed. Creative and intuitively appealing procedures began to be offered by computationally sophisticated bioinformaticians, but often had no strong statistical basis. A third emerging phase recognizes that numerical information produced by microarrays is not inherently different than other numerical measurements of random variables. Though new procedures tailored to the nuances of microarray data are called for, all the same laws of probability and basic statistical principles still apply. Our goal in this project is to develop and disseminate rigourously developed and tested methods for the design, analysis and interpretation of microarray experiments.

We are also interested in developing collaborative relationships with investigators using microarrays to study plants.

David Allison, Ph.D. (PI)
Grier Page, Ph.D. (Co-PI)
Stephen Barnes, Ph.D. (Co-PI)