At A Glance:
DDS, Columbia University (1993).
General Practice Residency program at Catholic Medical Center (1994-1995).
Specialty training in pediatric dentistry at the University of Southern California (1997)
Began DDM in oral biology, with dental informatics fellowship, at Harvard School of Dental Medicine in 1998.
MS in medical informatics from MIT (2001) with a special interest in Bioinformatics. Thesis, “Reproducibility of mRNA Gene Expression Measurements Across Microarray Technologies.”
Current affiliations: Decision Systems Group, Brigham and Women’s Hospital Children’s Informatics Program, (directed by Isaac Kohane) in the Department of Genetics, Harvard Medical School; Department of Infection, Immunity and Oral Medicine of Harvard School of Dental Medicine; and Health Sciences and Technology of Harvard and MIT.
Co-author on 11 microarray-related publications currently accepted, in press, or published since 2001. Two more in preparation. http://web.mit.edu/wpkuo/www/publications.html.
Winston Patrick Kuo has co-authored more microarray papers in the last year than many researchers do in a decade. These publications range from ones addressing basic technology issues, which he has familiarized himself with in his work at Boston’s Children’s Hospital Informatics program, to those on DNA microarrays in dental research — his clinical specialty. Last spring, Kuo also drew attention for a poster comparing Affymetrix, Agilent, and Motorola (now Amersham) CodeLink microarrays. He is now preparing a final summation of that research.
While such work would be enough to keep an ordinary young researcher busy, Kuo also conducted a study last year with Tor-Kristian Jenssen of the Norwegian Radium Hospital, in Oslo, Norway, who was a visiting PhD student at Harvard. Kuo and Jenssen re-analyzed gene expression data from a microarray dataset on breast cancer to examine associations between gene expression patterns and patient survival. They published their results in a recent issue of Human Genetics (“Associations between gene expressions in breast cancer and patient survival,” (4-5):411-420).
Now, Kuo has moved on to Darwin’s finches, on which he is conducting gene expression studies to answer questions of classical evolution.
BioArray News recently caught up with Kuo, between hospital hours and a presentation at a medical informatics conference in Texas, to discuss his flurry of work, his Human Genetics paper, and his current project.
So how did you get from dental school to medical informatics and microarrays?
My research in my second year of pediatric dentistry residency involved patients with primary and secondary cleft lip and palate at Rancho Los Amigos Medical Center [in Los Angeles], though it was more clinical -based research. I got more interested from a genetic perspective after reading [the] literature and decided to pursue further education, which led me to Boston. I started working with microarrays from a class project at MIT in the spring of 2000. It started with the cross-platform comparison, which then became my thesis at MIT, where Isaac Kohane and Lucila Ohno-Machado were my thesis advisors. And, ever since, I have been working with them in many aspects.
So how have you incorporated microarrays into your work?
Currently, I am engaged in a large-scale systematic comparison of multiple microarray platforms at the Cepko Laboratory, [in the] Department of Genetics [at] Harvard Medical School, with the hope that it will help answer the practical question of which microarray platforms provide adequate performance for reliable gene expression studies. In addition, we will explore several novel approaches to transform the gene expression values from the various platforms to make them comparable.
From a biological perspective, I am interested in craniofacial development and malformations as well as diseases related to the oral cavity such as oral cancer. From craniofacial studies, we hope to identify candidate genes that will represent distinct signaling pathways and region-specific transcription factors with important developmental functions as well as understanding those involved in malformations in the cranium.
In terms of oral cancer, our ability to predict the biology of pre-malignant lesions has been limited by controversial clinical terms, inaccurate and subjective assessments due to lack of well-defined criteria for grading, and the lack of genotypic/phenotypic-based biomarkers. It is our hope that we will be able to identify a number of such specific molecular biomarkers that will improve our ability to diagnose, prevent, and treat this aggressive disease. In addition, this may provide new insights with regard to the biology of this and other neoplasms.
What led you to the recent Human Genetics paper, in which you analyze breast cancer microarray data?
I wanted to start looking at the data generated from microarrays from a biological perspective. At the time, we were interested to see whether we could use gene expression to predict outcome. The only dataset available at that time was the breast cancer one from the paper by Sorlie et al. (PNAS 2001 Sep 11;98(19):10869-74).
Why did you choose to re-analyze this data?
At that time — February 2001 — this was the largest publicly available microarray dataset associated with clinical information, particularly survival data. It was interesting to investigate how microarray data can predict survival.
And why did you use the log rank test for analysis of survival by each gene?
We used this method to assess the predictive value of each gene, and also the method has not been used in this particular context with microarray data. The log rank test was implemented in [the program] Matlab, which made it practically feasible to screen all the genes.
Why did you use multivariate regression analysis to derive the genes that were highest in predictive value?
We decided to use that analysis, since the log rank tests considered only one gene at a time, whereas the Cox regression makes it possible to see how different factors (genes) could be combined to improve the prediction. Briefly, the multivariate regression analysis combines factors that have independent predictive value.
Have you compared your results to those obtained by other groups doing expression profiling of breast cancer samples?
No, we have not, since we were not able to find similar microarray datasets with survival data.
What future directions are necessary to further develop microarrays as breast cancer screening devices?
This is not really a question of using microarrays for breast cancer screening but probably more about microarrays in general. A general answer would be to improve the microarray technology.
What are you doing now in terms of follow-up studies?
We are currently investigating other cancers and diseases.
So now, I understand you are doing some gene expression work with Darwin’s finches?
I was very fortunate to get involved in this “classic evolution” project. I am working with Arhat Abzhanov, a post-doctoral student in the Tabin Laboratory, which is the same laboratory where I work.
As you are aware, Darwin’s finches are a group of closely related birds inhabiting the Galapagos Islands. Despite their close phylogenetic relationship, Darwin’s finches have evolved a number of distinct cranial morphologies adapted for particular kinds of food due to extremely high selective pressure.
By comparing expression profiles of several species representing distinct morphologies such as the sharp-billed finch, Geospiza dificilis, which has a small symmetrical beak; large and medium cactus finches G. cornirostris and G. scandens, with long sharp beaks; and large, medium, and small ground finches, G. magnirostris, F. fortis and G. fuliginosa with very deep, thick but short beaks, we are hoping to identify genes that correlate well with particular morphologies, or more precisely, with different dimensions of beak morphology. We expect that these candidates will represent distinct signaling pathways and region-specific transcription factors with important developmental functions. We will confirm the candidates by looking at their expression patterns both in finches and in [the] chick, our other] bird model system.
The more long-term goal is to learn if our candidates perform important developmental functions by doing functional analyses — both mis-expressions and knock-out experiments. It would also be interesting to perform broad comparative analyses of our candidates among other species of birds and other vertebrates.
We have also started examining the cactus finches, G. cornirostris and G. scandens and [have] in process, the ground finches, G. magnirostris, F. fortis and G. fuliginosa. We are using in-house cDNA microarrays with about 22,000 cDNAs.
Finally, what would be your ideal microarray experiment, if you could design one?
In terms of arrays in general, having internal good controls is needed. In short, given the current data quality generated from the arrays, I would suggest that replication/reproducibility issues are important.