At A Glance
David Henderson, assistant professor of Animal Genetics, assistant professor of Biostatistics, University of Arizona, Tucson
1993 — BS, science and animal science, Texas A&M
1996 — MS, science in animal science, South Dakota State University
2001 — PhD, genetics, Virginia Tech
1995-1998 — Quantitative geneticist for PIC, Franklin, Ky.
University of Arizona professor and biostatistician Dave Henderson has spent the past five years on the design and analysis of microarray experiments in the agricultural field. He is the statistician associated with the National Science Foundation-funded maize microarray project. Earlier this year, Henderson and his colleague Bruce Walsh published a paper in the Journal of Animal Science reviewing methods of gene expression analysis in animal breeding research. This week, Henderson talked with BioArray News about his research and a biostatistician’s view of using microarrays.
How long have you been working with microarrays and how did you get started?
I have been working with arrays since 1999-2000, and it was through some consulting work at Virginia Tech. We were using a custom array for loblolly pine.
Tell us about the research that was published in the Journal of Animal Science.
This was an invited review article. It was part of a symposium that was held a year ago at an animal science meeting in Phoenix. The symposium had to do with new molecular genetics technologies and how we’re going to handle them quantitatively. After the symposium [my colleague, Bruce Walsh] and I wrote this article. We share this belief that we can use metabolic control analysis to explain the relationship between metabolites, proteins, and the expression of genes.
What did the research involve?
The research involved trying to find novel statistical methods to relate genes to each other — just to their expression values — with the idea that we can group them into metabolic pathways. And using these metabolic pathways we may be able to find genes that are instrumental in producing certain phenotypes. In animal genetics, we’re very interested in finding selection targets — genes that have some kind of polymorphism that we can directly select on or maybe there’s some kind of dosage effect. Maybe we can find animals that have better high expressers and select for those that may produce some desirable phenotype.
What tools did you use in doing this?
We used things like Bayesian belief networks, factor analysis, linear statistical models.
What are the implications of this research?
We’d be able to find novel gene targets that could be used to identify animals that produce superior phenotypes for economically important traits. That’s the direct result. The indirect result is that these methods could also be used to identify drug targets in plants, animals, and humans — and possibly genes that are involved with the onset of disease or are involved with protection against environmental factors such as heat and UV stress.
Are there commercial microarrays available for animal breeding purposes, or are people involved in this research making their own arrays?
There are not really any commercial arrays, other than an Affymetrix [bovine] array. Most of the arrays, though, are available through consortia, such as the Michigan State Consortium. There is one at Texas A&M also involving the University of Arizona and the University of Missouri. There are a couple of other smaller arrays available that we print here at Arizona. Certainly the largest one will be the Michigan State one.
What is the content of the arrays?
Most of the arrays are cDNA arrays. We’ve developed certain BAC libraries based upon taking tissue from certain organs — skin, liver, mammary, gastrointestinal tract, and muscle. Mammary is probably the largest tissue represented on these arrays because most of the researchers are dairy science oriented. Michigan State is the only group that has a long oligo array. We found unique sequences and then had Qiagen design some 70-mers, and you can purchase the set from Qiagen. I don’t know if you can do that directly or you have to go through Michigan State. We just purchase the array from Michigan State. But we do have the ability to print the cDNAs here at Arizona.
Have you received any commercial interest for the work you’ve done?
I have had some interest, mostly in my statistical methods, and that’s through a company in Siena, Italy, called Siena Biotech. They’re using some of my methods on their protein data. I have another commercial interest, and that’s with the Livestock Improvement Company in Hamilton, New Zealand. They’d like to map QLT for dairy traits, particularly QLT that are associated with expression values. They’re going to use about 1,500 arrays on their commercial dairy population, and we’re going to try to find chromosomal regions in the genome that are associated with regulation of gene expression.
What are some of the biggest obstacles you face in using microarrays in your research?
We really have to rely a lot on approximate methods for significance. Computational problems are the biggest limitation. Most of the time we find that our test statistics are not distributed asymptotically, which causes some problems for inference. We have to do things like bootstrapping, which is kind of computationally not feasible at the moment. I’m working on some Bayesian methods that may be able to get around this, but it’s not known if that will work. We’re still in the design phases of that.