At A Glance
MD, PhD from Duke University
Developed first spotted microarray system along with Pat Brown and Dari Shalon at Pat Browns lab at Stanford from 1990 to 1993.
Set up microarray core laboratory at UCLA and serves as faculty director. Current research uses high-density microarrays to conduct gene expression profiles of brain tumors and to correlate these gene expression patterns in to clinical outcomes in order to better define why some tumors respond to specific therapies and others do not.
QWhen you came to UCLA, how did you go about setting up the microarray lab?
AThe first thing we did was use the Pat Brown arraying technology (the home-made arrayer) to enable ourselves to create generic human and mouse arrays. Since then, we have helped people print up E. coli arrays and mouse arrays. We tried to make it a fairly inexpensive printing facility. Over the last year and a half or so, we brought in Affymetrix technology as well, because its a rather user-friendly platform, with the downside being that its more expensive than spotting your own arrays
QDo you find theres a difference in Affymetrix GeneChips, in terms of consistency and reproducibility?
AAffymetrix GeneChips have greater feature consistency than cDNA arrays, and the likelihood that different arrays have the same amount of probe is certainly greater than with our spotted arrays. But with cDNA arrays that are printed well, with surfaces and background that are in an acceptable range, you can generally get comparable results.
The benefit of the Affymetrix technology is that when you do an experiment now, it is very comparable to the experiment that you did six months ago, whereas with spotted arrays, there is not the same consistency
QAffymetrix, along with some industry observers, has predicted that microarray users will switch over from do-it-yourself cDNA arrays to pre-printed oligo arrays in the near future. What do you think?
AI think it has to do with price. If the folks who can do in situ synthesis of oligos Affymetrix, Agilent, or [Madison, Wis. startup] NimbleGen can do it at a very low price with tens of thousands of array probes on the slides, and have oligos that are sufficiently long and sufficiently specific, then our spotted arrays will have a very difficult time keeping up. Right now, the advantage is our spotted arrays are a lot cheaper than these commercial products.
QWhy do you need longer oligos?
AThere are issues with how the 25-mer oligos behave on the Affymetrix arrays. It is conceivable that data is going to come out cleaner with longer oligos, but the jury is still out on that one. It would be nice to see someone do a side-by-side comparison of Agilent [oligo] arrays and Affymetrix arrays to see which performs better.
QHow do you address the often thorny problem of microarray data analysis?
AIt depends on what sort of problem or question we have in the lab. If we are looking at tumor types that are distinct, clustering gives us global information that the expression profiles are looking informative. Then we use a wide variety of statistical approaches to look for the subset of genes that most contributes to class distinctions. Sometimes we use principal component analysis to find the genes that are contributing most to what we are studying.
We are generally very lenient on the first pass through, and we accept information that does not robustly predict a condition. A lot of low level expression genes may be the most interesting genes. If something is reproducibly coming out even as a one and a half-fold difference, I wouldnt ignore that information. Sometimes the differences might be a hundred fold in reality. Microarrys tend to underestimate differences in expression.
QWhat do you think is needed most in the near future to improve microarray research?
AOne thing that we are constantly discussing is having standards in the field so people might be able to compare their data better. We ought to have a set of universal RNA samples. So if somebody in Salt Lake City is starting up an array facility and wants to know if the technician has mastered the array hybridization procedures, the person can use this universal reference RNA, and then go to the universal reference database to see if they have gotten the same answer as the standard answer.