At A Glance
- Postdoctoral researcher since 2000 in the laboratory of Jeffrey A. Johnson, assistant professor in the Pharmaceutical Sciences Division, University of Wisconsin-Madison School of Pharmacy.
- Li received his PhD from a joint research program between the National Key Lab for Virology and Genetic Engineeringy, Beijing, China, and the Fourth Military Medical University in X’ian, China.
- Lab’s current interests: “To discover ways to increase the defense mechanisms in [the] brain by activating multiple antioxidant defense genes simultaneously through activation of the antioxidant response element (ARE).”
- Li and Johnson co-authored several recent papers: “Differential gene expression patterns revealed by oligonucleotide versus long cDNA arrays,” in the October issue of Toxicological Sciences (69, 383-390); “Time-dependent changes in ARE-driven gene expression by use of a noise-filtering process for microarray data,” Physiological Genomics, 9(3):137-44); and"Microarray analysis reveals an antioxidant responsive element-driven gene set involved in conferring protection from an oxidative stress-induced apoptosis in IMR-32 cells."(Journal of Biological Chemistry, 277(1):388-394).
So you and professor Jeffrey Johnson just published an article in Toxicological Sciences comparing cDNA arrays to Affymetrix chips. You compared the performance of Incyte’s UniGem v 2.0 cDNA chips to the Affymetrix U95Av2 array by hybridizing the same sample - a neuroblastoma cell line treated with the antioxidant metabolite tert-butylhydroquinone (tBHQ) - to both chips then examining how good each one was at detecting genes known to be previously upregulated by treatment with thBHQ. What led you to do this experiment?
Two years ago, when I first came here to work as a postdoc, Jeff had already started sending samples to Incyte Genomics. Incyte provided the service to make the cDNAs, label them and run microarrays. But every time we would find that some of the genes which [we] had already confirmed should be changed, and do change in the system, were not. We questioned whether something was wrong with our samples. So we also sent our [samples] to Affymetrix in California. Their preparation was quite different. When we got feedback, it was very good. A lot of genes that should have been changed showed up in the data list [as changed.] We were very excited about this. We wanted to know why Incyte chips had data problems.
Affymetrix was also interested in this, and they provided our lab a lot of free chips. We sent this sample to Incyte — it’s like single blind research. We ran almost 20 Affymetrix arrays to confirm [our result]. A lot of academics labs have no money to do this comparison. We also paid $10,000 to run several pairs of samples in the Incyte system. With Affymetrix we finally bought the whole system, so we run the data our selves.
So did you then go present this information to Incyte?
We figured out why Incyte had a problem in late fall 2001. We still wanted to send another sample to Incyte, but Incyte shut down their system. We also traced [from] their paper, published in Nucleic Acids Research that their system had problems. They had a lot of cross hybridization in their arrays. [Editor’s note: See article by Incyte scientists in Nucleic Acids Research 2001 Oct 15; 29(20):4251-6, “DNA exhibits multi-stranded binding recognition on glass microarrays.” Also see related article by Incyte researchers in Biotechniques, 2001 Nov; 31(5):1182, 1184, 1186 “Hybridization cross-reactivity within homologous gene families on glass cDNA microarrays.”]
Nature had news talking about [how] Incyte shut down their microarray [service] system because they transferred to proteomics and genomic databases. The real reason, we believe is that their [microarray] product may not be as good as Affymetrix’s. Affymetrix now controls over 50 percent of global screening arrays sales even though their product is so expensive, which is a testament to their accuracy.
Since Incyte left the business a while ago, what relevance do your results have now?
A lot of the other companies like BD Biosciences Clontech are gradually changing from long cDNA arrays to oligo arrays. BD Clontech [in] their membrane arrays is still using long cDNAs. Long cDNA arrays have a lot of problems. The cDNA ranges from 500 - 5,000 bp, averaging 1,000 bp This kind of cDNA can bind to a lot of things. Inconsistent Tm value among probes is also a problem for hybridization. Overall, long cDNA arrays are subject to a lot of cross-hybridization
The two-dye system [Incyte used] also has problems. For example, different fluorescence-labeled nucleotides may be incorporated with varying efficiency, altering the ratio as a result of enzymatic contributions rather than transcript abundance. Next, multiple experiment comparisons are not possible without replicating the reference sample, which in some cases may be difficult to obtain. Spectral overlap between dyes can also complicate interpretation of algorithms used in analysis. Lastly, executing signal amplification schemes in two colors is more complex than ina single-color because multiple haptens are required.
A key assumption when employing microarrays to profile gene expression changes is that the quantified signal intensities are in linear relation to the real expression levels of the corresponding genes. Accumulated evidence indicates that data generated from oligonucleotide microarrays displays a good correlation to SAGE (Serial Analysis of Gene Expression), RT-PCR, and Northern blots. In a cDNA array, the nonlinearities were revealed by serial dilution experiments. With Affymetrix, synthesis of oligonucleotides by photolithography offers the advantage of abolishing the need to hydrolyze the oligonucleotide from its synthetic support and reattach it to the microarray. They have 14-20 probes representing one gene. Affymetrix also designed the mismatch which may presumably represent the cross-hybridization.
Have you considered other arrays, other than Incyte and Affymetrix?
BD and Operon are making long oligos with 50 to 70 base -pairs. I think that the future is in long oligo arrays, because they are cheaper and they are still reliable. We are going to test these [long oligo arrays] to make sure they are working as well as Affymetrix. Their price is still high - even higher than Affymetrix. I believe that the money they spend on manufacturing is much lower than Affy, because they have just one oligo representing one gene without the corresponding mismatch probes.
What do you think would be needed to make a product better than Affymetrix’s chips?
The one that can beat out Affymetrix is one with decreased manufacturing costs, like those of NimbleGen of Madison, Wisconsin. Affy uses a lithographic mask. Nimblegen has a system that can synthesize probes in situ without using the mask, but using a changing mirror, so that you can decrease the manufacturing costs.
The second change is that we talk about perfect match and mismatch [with Affymetrix.] A lot of papers published say that the mismatch is not a real representation of cross-hybridization. So you do not need a mismatch. You can just call the perfect match. The software, dChip [http://www.biostat.harvard.edu/complab/dchip/] just analyzes the perfect match.
What kinds of software do you use to analyze your data?
We are using the original software provided by Affymetrix. We also use other software, which you can download. We modified it slightly when the need arose. (See our paper in Physiological Genomics). We introduced rank analysis. It’s a simple way to help a lot of biological researchers who are not familiar with biostatistics.
What do you think of the newer Affymetrix software compared to Version 4?
We are doing the comparison for this now. We do not know which one is better. We need a lot of confirmation. But I will tell you one thing: [It depends] whether your system is tissue or cell line. With the cell line the data from version 4 and version five match more. If [your sample is] from tissue the have a bigger bias. The gene sets generated have some overlap but not as much as the cell line, which is very pure, very consistent.
What is microarray research like in China and how does it differ from that in the US?
There is more money to provide for research in the U.S. A lot of Chinese labs are still using the Incyte format-the long cDNA, home-made arrays. This decreases the quality of your products. They are not very reliable or reproducible. If you dot the array yourself, sometimes it is good, sometimes not. As far as I know, several companies and universities in China are working on this, but their products cannot compete with [those of] American companies and the publications are limited.
So does this mean that the Chinese market is open for the likes of Affymetrix?
If Affymetrix can decrease their prices, I believe they can open the market in China. But even some American labs still cannot afford them. Other companies also have very competitive products,and promise customers that their product can be reused, while Affymetrix does not allow the customer to do so. At present, arrays are not like PCR. PCR was first introduced in the late 1980s, and the thermocycler became widespread, but nowadays, PCR is so cheap and so routine. I believe the microarray will be the same way in the future. But it takes several years to make the prices go down.