Name: Edison Liu
Title: Executive Director, Genome Institute of Singapore
Professional background: 2001-present, executive director, Genome Institute of Singapore; 1996-2001, director, division of clinical sciences, National Cancer Institute; 1993-1995, associate professor, departments of medicine and epidemiology and biochemistry, University of North Carolina at Chapel Hill; 1987-1993, assistant professor in medicine and oncology, University of North Carolina at Chapel Hill
Education: 1983-1987, postdoc, department of microbiology, University of California at San Francisco; 1982-1985, hematology fellowship, University of California San Francisco, Moffitt Hospital; 1980-1982, oncology fellowship, Stanford University; 1979-1980, residency, Barnes Hospital, Washington University, St. Louis; 1978, MD, Stanford University; 1973, BS, chemistry, Stanford University
SINGAPORE — Permanence kills creativity, according to Edison Liu. It's the personal philosophy that drove him from a full-time professorship at University of North Carolina Chapel Hill, through a permanent position at the National Cancer Institute in Bethesda, Md., and on to a contract position as executive director of the Genome Institute of Singapore within a period of six years between 1995 and 2001.
Now, with close to a decade of experience running GIS under his belt, Liu is eager to take on more projects. He was re-elected last month to serve as president of the Human Genome Organization, a position he has held since 2007. In particular, Liu sees HUGO, which is now run from Singapore, as a vehicle to cement the role of developing Asian countries in genomics and to broker deals between West and East that not only foster the application of genomics in medicine but also spur economic growth.
An early result of this effort appeared in Science in December 2009, when HUGO's Pan-Asian SNP Consortium reported on genetic patterns in more than 70 Asian populations. The HUGO team, which included researchers from 10 Asian countries as well as investigators from the US, used Affymetrix microarrays to map the genetics of 73 Asian populations. According to Liu, the consortium is planning a follow-on study that will be at least twice as large and could potentially include Middle Eastern and Amerindian populations.
To keep abreast of these endeavors and to learn how he views the evolving approaches to genomics research, BioArray News spoke with Liu at his office in Singapore's Biopolis last week. Below is an edited transcript of that interview.
You were recently re-elected to lead HUGO. How has the organization changed under your leadership?
HUGO is a very unusual organization. It was structured originally as an elite organization of the greatest thinkers in planning the human genome project. It was in many ways an attempt to internationalize the human genome effort with Europe, the US, and Japan playing the major roles in that effort. As time went on, HUGO, which sought to have a conciliatory role as a venue for discussion and exchange of ideas, took on issues around nomenclature and ethics. But in each of the individual countries, as they matured, particularly in the US where [the National Human Genome Research Institute] was set up for that purpose, [those local institutes] began to take the intellectual leadership. Of course, that is where the money was, and where the money goes is where people head.
When the genome was [published in] 2000, then the raison d'être for the organization came into question. It was an organization that was structured to help sequence the human genome. Now it was done; what was it going to do? So for a period of time, until about 2005, things were just business as usual. In 2005, I started to really get involved in HUGO. There was a lot more questioning of where we were going to go. And out of that discussion in 2007 I was elected and I think the selection was in many ways a reflection of the geographical shifts taking place in science, and secondarily a bit of the enunciation of what I thought HUGO should be.
When I talk about the geographical shifts, I don't mean shifts of greatness in any way. I think the center of gravity is more equalized between the West and the East. It was intriguing for me, because we had a HUGO Asia Pacific meeting here in 2004, and immediately after that was a HUGO meeting in Berlin, and we had more people in Singapore for a subsidiary meeting than there were in Berlin. That was clearly a signal that the Asians were very interested in this field for many reasons and that the shift in enthusiasm for the role that genomics plays in human health issues was normalizing from the first world to the emerging countries like China, India, Korea, Singapore, and Taiwan.
[ pagebreak ]
For the second part, what I was suggesting is that perhaps our strongest suit is to ask what genomics can do for humanity at this point — not just sequencing the human genome, but determining the consequences of that sequence for genomic medicine. Another part was to state the reality that many countries were emerging from third world status to first world status. These countries all view genomics as a direct path to high-impact biological discovery. A great example was the work of Andy Simpson and his colleagues in the early 2000s where, out of nowhere, São Paulo became a force in genomics because Brazil got together and sequenced some plant pathogens that were pertinent to economic issues, which spawned a whole generation of computational biologists because there was a substrate to work on. There was actual national pride that came with that.
Since 2004, we also began working with the Pan-Asian SNP Initiative. It was the HUGO Asia Pacific group that got together and decided they wanted to do something. It was the meeting here that actually launched it. Over several years, we mounted the resources, a way of doing business with each other, and then came up with that Science paper. That was done because we all believed in the technologies, but we also believed in Asia as an entity where there is not only the potential for scientific excellence, but there is also the uniqueness that we bring to the table. So, this all kind of converged.
What are your current plans for the organization?
Well, first of all, HUGO needs to be an organization that supports rational discourse around the implications of the technologies and society and humanity. We want to be brokers for technology transfer between those who know and those who want to know. We want to enunciate how countries can use this technology for both economic and intellectual development as well as, of course, health. The health part is obvious, but the economic development part is less so. We believe this technology will drive the incursion of biology into all aspects of the economic cycle. We want to be the facilitator for international collaboration and finally to assist emerging countries to achieve scientific greatness through genomic approaches. These approaches are now integrated into the fabric of biological and medical investigations. That in a nutshell is what we want to do.
On a practical level, we broker these projects. We are not a funding agency, but by nature of the bully pulpit we have, we can draw people together and facilitate companies to support these projects and funding agencies to support these projects. These projects are cross-national. We want to enhance our meeting. Finally, the new HUGO Journal is another vehicle for us. It is in production, and we expect our first issue to debut probably in June. We want to focus on genomics, genomic medicine, genomic technologies, and policy as well.
You mentioned the Science paper and the Pan-Asia SNP Initiative. What's the next step for that consortium?
We have the [annual Human Genome Meeting] coming up [in Montpellier, France, in May] and we'll have another meeting for Pan-Asia SNP 2, and we hope to double the size of the population, and to include indigenous populations from Australia and New Zealand. We have discussions with Amerindian groups in South America and Latin America. Next year's [Human Genome] Meeting is in Dubai, [United Arab Emirates,] and so we are talking to the Center for Arab Genomic Studies and could then extend the reach of the study. We will increase the comprehensiveness of the search. Instead of 50K, it will be 600K plus with CNVs included in that process and with other targeted SNP searches. The hope is that it will be much more refined and much more useful in terms of what we hope to accomplish.
What did you use for the first part of the study?
The Affymetrix GeneChip Human Mapping 50K Xba Array. It's pretty rudimentary nowadays, but when we first started in 2004, it was still being used. Most of the work was done before 2006. We spent about a year doing the analysis because we had to meet where we could, et cetera. And then it took us over a year and a half to get it published.
What would you use for the next phase?
I want to be vendor agnostic, but it would be one of the platforms that is quite comprehensive, with over 600,000 SNPs, that would also cover CNVs.
[ pagebreak ]
This study is not a genome-wide association study.
No, because we aren't associating anything. We are actually doing a survey. See, one of the things about having good arrays that are very dense is that you can do imputations to SNPs that are found by other means that [are] in linkage disequilibrium and make some inferences about frequency and all this other stuff.
We were hoping to infer haplotypes as well, but with 50K you certainly can't. With the 1M, you can make some haplotype imputations, depending on the structure within the population. Whether we will include trios to get at those haplotypes, at this point, I don't know. We are still in discussion. The original concept of this study was to make it easier for people to participate. So the question had to be narrowly focused on a migration or diversity question and not on disease association. In retrospect, that was the right move because it would never have got off the ground if we were to look at disease association. It's just way too hard to look across national borders.
Different nations have different capabilities. The most difficult is tissue accession and phenotype. It's much easier to go to a community and ask 10 members to contribute their DNA without capturing their disease state.
I read an interview with you in Nature about the Pan-Asian SNP Initiative. You described how Asian countries often see each other as competitors, and that this initiative brought them together. Why is that?
I think this may have been a first. First of all, there is a lot of recent history in the sense that people still remember the numerous wars in the whole region and wounds that still haven't healed. On top of that, every country in this region is based on export to the US and the EU. And so, consequently, every country views itself as a competitor to the next. The mindset has been historically very competitive. Now, that's changing with China and India also getting much stronger.
In the first project, you looked at 2,000 samples. How large do you expect this second project to be?
Probably around 4,000. Maybe 5,000. It just depends on the geographic domain and we hope that now the technology is so widespread and the quality control standards are pretty set that we can actually assess whether a data set is usable or not. In 2004 when we first started that was the beginning of widely utilized standardized SNP analyses that could be generally used.
Where do you stand in the debate that is going on about the future of GWAS? Some predict a second round of GWAS based on arrays, and then there are others who say that smaller, sequencing-based approaches might be more fruitful.
Personally, it's like asking, 'Do you like French food or do you like Chinese food?' I mean, they are both great except they are really different. The truth of the matter is that you need both, and the reality is they are not separate; there is a continuum between what is common and what is rare and the question is what threshold you choose. I do believe that we are going to find more genetic causes of disease. They may be rare, they may be private, and what I mean by that is that it may be a private mutation in a gene that's common. In cancer, we are finding that all the time. So I suspect that will be the case in heritable disorders. The difference is that in cancer there is a phenotype that is pretty definable; you just need to sequence the genome. But there is almost no question that we might find more. But my belief is that five years from now, we won't be calling them GWAS or sequencing studies. It will all be one study. One will be shorthand and the other will be longhand. The cost of GWAS is dropping precipitously, so the limiting factor is not the GWAS, it is the extraction, storage, management of the DNA, and phenotyping. That is the major limiting factor.
What about funding? Are agencies still interested in supporting these projects?
JJ Liu here has really been helping to enhance the GWAS framework in China. There was a New England Journal of Medicine paper on leprosy and a Nature Genetics paper on psoriasis and an American Journal of Human Genetics paper on the stratification of diversity in Chinese populations on a geographical basis. All of these are pointing to the fact that Asians do have a different set of susceptibilities. So the Chinese are putting more and more money into this process, because they know they can mount very significant studies within a short period of time. No other country could mount something like that leprosy study. India has the numbers but it doesn't have the infrastructure to get the samples, organize it, and get the genotyping done.
[ pagebreak ]
If I were to hire you to consult me on my GWAS study, what would your advice be?
That would be dependent on the project, but let me just tell you what I believe is happening. I wouldn't undergo a GWAS study unless I had a definitive endpoint, a phenotype I could unequivocally ascertain. The second issue here is that, quite frankly, why do you want to do it? What is the endpoint? I say this because there are times when people do a study because the technology is available as opposed to doing a study with a definitive question at the end of the day. But what is the next generation of studies? There is almost no question that targeted sequencing is the next phase. There, the question is, 'What are the rare variants in critical regions of the genome and can we exploit that knowledge in highly defined populations?' That is where I think some of the most interesting observations are going to be.
But the same advice could be given for sequencing-based studies.
Absolutely. And, by the way, there really is no difference between these two approaches in mind. The only difference is that there are inferences you can make on base changes that are definitive and private that you can't make with the common SNPs. But I think that sequencing will make the greatest impact in linkage analysis, where you go back to the pedigrees in families and look at the types of mutations that get disseminated in these populations. We are doing some studies like that now. Still, sequencing whole genomes is expensive and you get this data where you don't know how to handle the whole data set. So we have been targeting the type of sequencing we want to the question.
When Illumina debuted its HiSeq 2000 system, several Wall Street analysts downgraded Affymetrix's stock because they saw a link between the availability of Illumina's HiSeq and [a potential drop in] Affy's microarray business.
That's what I am trying to say: arrays are like shorthand for sequencing. To be frank, they downgraded the stock because HiSeq will drive the cost of arrays down.
For us, it's not that bad. These arrays do add value and Affy does have a technology that is quite cost effective in terms of manufacturing. With their new chemistries, they have improved dramatically their call rates and the ability to look at certain loci. With those things in common, their precision coupled with their ability to drop production costs, what we are seeing is the same cycle we saw with IT chips. They are going into innovation and cost reduction as a way to be competitive. They will hit a certain cost structure whereby unless there is an alternative technology, a lot of companies will not get to the cost range that Affy can drop to.
We have played around with the GeneTitan system; it's fabulous. One can do multiplexing with high precision in a manner that is quite impressive. So it becomes a numbers game. Affy can go up to 2 million SNPs on a chip. That's getting a lot of territory on one product. In general, I don't think the array market will go away. It's not a dinosaur, but it is past its economic pinnacle in terms of revenues per unit. It's not going to return to the level of a few years ago.
Moving from genotyping to expression, what do you think about the options for researchers who want to do expression profiling today? If you were a young PI, what would you be using?
To date, the problem is that we do not have a normalized comparison from sample to sample on the sequencing platform. Data from expression arrays have gone through statistical validation and normalization so when you cut across the sample sets from different labs, you can extract some pretty interesting information. We don't have that well set up in sequencing at this point. It depends what your project is. If you want precision, then I would do sequencing. If you wanted to do relative expression, we don't have enough experience for it.
If you want to talk about splice variants, algorithms today are not that good at looking at splice joints. You have to deal with it and resconstruct it and it gets more and more complex. If one wants to look at new genes, then sequencing is the way to go. But if you are scanning to monitor a process and you are not relying on absolute precision, you don't need sequencing.
But arrays have their advantages too. I'll give you an example. The [Singapore] Ministry of Health asked us to look at H1N1. We have four sequencing platforms here, so we immediately started to do sequencing. We started with Roche 454 and continued with Illumina using nasal swabs. But we realized that by using these random primer sets, we were getting all sorts of human RNA and bits of virus. In one case we even had dog RNA because the person had probably kissed their dog on the way out the house. We got everything. At the same time, we constructed an array with Roche NimbleGen. We did some optimization. It took us two weeks to structure and to get the manufacturing to take place. We waited two months to get a set of samples to test. We developed websites to upload the data. This array is now being produced and sold by NimbleGen. So the array gave us all the information we needed and none of the junk. For that very specific purpose, it was perfect.
This question is the whole problem that engineers face with overengineering. If you have a problem, you can build an amazing structure around it. People will spend an enormous amount of money for a problem that does not deserve a Cadillac treatment. That is why I think that arrays will be commoditized to a certain degree, but manufacturers are still doing quite well. They just need to innovate on a continuous basis.
[ pagebreak ]
Where do you think array technology could go?
If I were a vendor, I would certainly focus on price point, speed, and then integration, and by integration I mean integrating DNA extraction and storage with the array analysis. That would be the lowest hanging fruit for what you might want to do.
You have been at GIS since 2001. What will the institute look like 10 years from now?
I left an endowed chair and full professorship in a state institution to go to the NIH for a full-time permanent position and left that for a contract position here. It reflects my philosophy that permanence kills creativity. That being said, you need to have a cushion, so that you are not flopping in the breeze, fearful, trying to feed your family. But as long as that is there, then to be fluid and to be challenged is the most exciting thing around. That is a long way of saying that I don't know where GIS will be 10 years from now, but I hope it won't be the same, because if it is, then it will be truly irrelevant.
GIS has morphed three times over since we arrived, and it's had to do with the fact that we had new personnel with new technology and chasing after the most important questions around. As different talent comes to GIS, we sculpt our direction to suit that collective talent. You are not going to run marathons with Olympic swimmers. You might as well find a body of water and beat everyone else. That is what we have been trying to do.
What it has also morphed into is systems biology. When we started out, we were heavily into technologies. A full 40 percent of our people were technology oriented. Now it's 15 percent. The largest group is computational biologists. They are two-thirds of our institute. It reflects the changes in technology and the changes in the institute. Going forward, I think that the systems reconstruction, the systems biology approach is going to be big, that we will all become more computational in nature. A decade from now, we might not be an institute for genomics. We could be an institute for integrated systems biology. That is probably where it is going to go, if I could predict where it goes.
I think we will also spin off more and more units that will assist in development of startups and impactful activities for national economic development framework. And this ties into the fact that emerging countries are not just viewing genomics as a way to improve human health, as the US is, but also for economic development. Similarly, I think what we are going to be doing will be expanding beyond the medical remit that we had earlier on. We had two of our scientists convert completely from yeast biology into algae genetics for biofuels. It's really exciting.