Research scientist in protein microarrays product development
At A Glance
Name: Dorothy Yang
Position: Research scientist in protein microarrays product development, Agilent Laboratories, since Jan. 2001.
Background: Senior scientist in assay and product research, LJL Biosystems, 1998-2000.
Research associate, University of Connecticut, department of chemistry, 1994-1998.
Postdoc, Institute of Organic Chemistry, University Bern, Switzerland, 1992-2993.
PhD, University Libre de Bruxelles, Belgium, 1992.
BSc, Peking University, China, 1988.
Last month, Dorothy Yang gave a talk at the Cambridge Healthtech Institute's Beyond Genome conference on a protein array she is developing for Agilent Technologies (see ProteoMonitor 6/24/2005). ProteoMonitor caught up with Yang to find out more about her background and Agilent's protein array platform.
How did you get into proteomic work and developing protein arrays?
Actually, it took me a long time to land in the field of proteomics. My original background is physical chemistry. I graduated from Peking University with a bachelor's degree in 1988, majoring in organic chemistry. And I got a PhD from the Free University of Brussels in Belgium, working on methodology development. At that time I was more focused on the effect of ultrasound on several types of chemical reactions. In mid-1992, after I finished my PhD, I went to Switzerland as a postdoc in Bern University. And I started working on electroactive compounds. From that time on, I started working on large, conjugated organic systems, especially molecules with very interesting photochemical and electrochemical properties. That was the turning point of my career. I switched from physical chemistry to photochemical molecules, to organic dyes and organic fluorophores. So in early 1994, I joined the department of chemistry at the University of Connecticut as a research associate, and our research focused on photosensitized reactions using organic dyes. Basically, the electron transfer reaction between organic dyes to organic substrates. Beyond that, I helped Dr. Gary Epling manage his lab on a daily basis. So this had nothing to do with proteomics.
In 1998, I joined LJL Biosystems, which got acquired by Molecular Devices around mid-2000. So 1998 is the time that combinatory chemistry reached its pinnacle, and libraries of billions of organic entities needed to be screened. And together with other scientists, we started LJL Biosystems' reagents business. Our job at that time was to develop reagents for high-throughput drug screening. My job, in particular, was to develop long-lifetime and long-wavelength fluorophores, and conjugate these fluorophores to proteins and specific peptides. A good thing for me was that I worked very closely with the development group there, so I had the opportunity to learn a whole new different world from the biochemistry based on the material I made. It was really a gratifying time for me, and I started to appreciate the power of high throughput. I joined Agilent lab in mid-2000. I was hired actually by Agilent at that time to join a large team to develop a point-of-care device. What we were trying to do was measure several parameters in blood for patients with diabetes and cardiovascular conditions. My job was to develop a sensing mechanism for glucose and oxygen and carbon dioxide based on long-lifetime fluorphores, and incorporate them into the sensing matrix embedded in the device.
Shortly after I arrived at Agilent labs, Agilent decided to divest the medical division to Phillips, so the project kind-of dissolved. And I was transferred to the life sciences department to work on protein arrays. So that was a long road for me to get into proteomics. It was about 2001 when I started working on the protein array program. I started to develop surfaces for protein arrays.
How did your background, in terms of protein research, antibodies, and fluorophores help you to develop the surfaces for the arrays?
Actually, I think the most important thing for protein arrays is the substrate. The substrate has to give you the lowest background, which is the starting point of any kind of assay development. So we manipulated many chemistries on the surface, and the goal was to have the lowest background and highest signal possible. And once we decided what type of surface we were going to use for the protein array, I started off doing a lot of assay development with other chemists and biochemists in the team. So I think my background helped a lot in terms of structure and property relationships. That's how organic chemists are trained we first understand structure, and we kind of estimate property. So I, together with other team members, started to develop the protein array's basic protocol. And for me, I learned throughout the program as well, either through reading or taking classes on protein chemistry and biology and cell biology. So I learned a lot beyond my organic background.
Were you working directly on an assay protocol for hepatocellular carcinoma?
Actually, when we started developing the protocol, we used a model system. The system is much simpler than the human samples we receive, because all those human samples are very precious. That's another advantage of protein microarrays, because in real protein arrays, we just use a tiny bit of sample, and we can measure many parameters. When we developed the protocol, we used a model system that is a mixture of several purified antigens. Once we set up one protocol, and we started to buy some serum samples from commercial sources, we used serum sample to verify the protocol and to determine if it is suitable. Once we used the serum sample, we set up the protocol, and we started to assay real, precious human samples from Dr. So [the director of the Asian Liver Center at Stanford University], and from other collaborators as well.
What were some of the difficulties in developing the array?
One difficulty, I think, is to measure multiple proteins all at the same time. We had to find the condition that is suitable for most of the parameters we are measuring. That means, we needed to find the unique condition that is suitable for most of the antibodies on the array. So we started out with the blocking buffer which blocking buffer will give us the best signal and lowest background? And the binding time we did two hours, four hours, overnight. And the temperature room temperature, four degrees. So all those kinds of experiments we played around with. And the rest of it is just tedious work you have to try out each condition and see which eventually comes out the best.
Aside from optimizing conditions, were there other difficulties in terms of the content of the array?
The content of the array is inflammation based. If you look at our antibodies, they are all inflammation-related proteins. And what we're trying to do with this inflammation array is to have a general picture about the inflammation process. For example, we work with our collaborators to decide which are the proteins they are interested in. So they give us feedback in terms of the inflammation process what are the proteins they are interested in. And our job is to put them together and print antibodies on the array.
The composition of our array we have many families of molecules involved in different processes throughout the inflammation picture. So we are trying to capture a very large picture of inflammation. And when you look at disease either cancer, or particularly liver cancer when the cancer starts to grow, your body even initially wants to fight it off, and that's the inflammation process. The same thing holds true for cardiovascular disease. So before the plaque ruptures, there's an inflammation response. And of course with autoimmune disease, where we're working with Bill Robinson, it is the same thing.
So I think I can say our array has some generalized applicability to many types of diseases, if we can capture the picture right.
Do you think the content for a liver cancer array can be used for a cardiovascular array?
I can't say the same content, but what I can say is there may be some overlap of antibodies. Of course each disease has a different mechanism, but looking at it from an inflammation point of view, there may be some overlap.
What we've found out is even if it's the same array, we can distinguish the patient with disease from control for several different types of diseases. And I think we're right. Just looking at inflammation, different types of diseases may have proteins overlapping.
When was the hepatocellular carcinoma array finished?
We've only done a very small study, which gives us very preliminary results. At that time [of the study], we had only 13 cancer patient samples and 10 control samples. And we could distinguish the patient group from the control group. We only missed one patient. But this is such a small set, so I can only say it's preliminary. For the next step, we're going to assist in doing a similar study, but with a much larger sample set. We're trying to evaluate the preliminary findings.
What would be the next step after the array is validated? Would you try to develop it into a test?
In Dr. So's group, yes. He wanted to find some biomarkers that he could use to do early diagnosis of liver cancer, because this is key for survival. And what we hope with protein arrays is that we can find a series of biomarkers that can distinguish patient from control. With this series, they are going to validate with an extremely large set of samples thousands of samples. They will use this group of proteins and validate, and eventually select maybe several, together with the current biomarker, alpha-fetoprotein, as a diagnostic. Because the problem with liver cancer is that AFP is the single biomarker now, and it will miss 50 percent of the patients.
Even from the small set of samples they gave us, half of those patients' AFP levels were normal. So that means AFP is not a very specific biomarker for liver cancer.
We have several proteins that can be used to distinguish patients with cancer from control. But the most difficult step is that it has to survive the multiple center testing. The same biomarker has to survive through the very rigorous validation process by different medical centers and clinical laboratories if you really want to put it in a diagnostic market. It's a very long-term project.
How many proteins are on your array?
Currently, 65 antibodies. About 10 to15 were used for distinguishing liver cancer. For these 10 to 15 proteins, the expression level is different between the control group and patient group. Our array platform is very flexible. We can add or eliminate any time we want, which gives us a lot of freedom.
Do you have plans to market this array?
Currently there is no commercialization plan. It's just a research project. At Agilent labs, there are many projects that involve working on technologies related to the business division, and protein arrays is just one of them. Currently, we don't have any commercialization plans for it yet.
Do you have any plans to develop some arrays for other diseases?
Currently, we have several other collaborators who are doing cardiovascular disease research. We are collaborating them. Also with Bill Robinson's group at Stanford University which is doing autoimmune diseases.
With different diseases, you have different groups of interesting proteins, but there are some overlaps. From a protein array point of view, we can just print one generic array and for different diseases, we can choose different detection antibodies to selectively detect different antigens.
Protein arrays maybe one day will be very important in the medical research community. From Agilent lab's point of view, we have to get the technology ready.
What are you working on for the future?
For the foreeable future, I think I'm going to do biomolecule detection. At Agilent labs, we are always looking at other technologies that increase the sensitivity and selectivity. But we can always use protein arrays as a benchmark to do multiplexed protein detection. This is my personal interest as well as my passion, so I will work on that protein detection, or biomolecule detection in general.