At A Glance
Assistant professor, biostatistics, Mailman School of Public Health, Columbia University
2000 – PHD, Biostatistics, University of Washington
1993 – BA, mathematics; MA, statistics, Boston University
Manisha Desai, 33, is an assistant professor of biostatistics at Columbia University and the organizer of a semi-monthly workshop on microarrays, which was started in May 2002. The workshop draws a dozen or so microarray-related researchers from within the university for a 90-minute session about microarrays, from soup to nuts.
Though she might not like the description, Desai has been a microarray statistics evangelist since shortly after her arrival at the biostatistics department of Columbia University’s Mailman School of Public Health in 2000. The workshop, she hopes, can one day become a class, giving the school’s 90 graduate and doctoral students in biostatistics exposure to genomic data and entrée to employers interested in those skills.
Desai is also part of what could be described as a microarray marriage: Her husband Adam Olshen is a faculty member in the Biostatistics Department at Memorial Sloan-Kettering Cancer Research Center, concentrating on microarray statistics.
BioArray News recently spoke with Desai about the Columbia microarray workshop and her hopes for the future.
How did you get involved in microarrays?
I didn’t get involved in genomics until I began my dissertation and then, it was completely different data than gene expression. I was working with loss of heterozygosity data, examining data from allelotype studies designed with the goal of finding tumor suppressor genes. So that wasn’t high-throughput data, but it was what sparked my interest in genomics.
I came to Columbia in 2000 and started my job here right after graduation. At that point, there were a lot of statisticians jumping on the microarray bandwagon. Our biostat conferences had become something like 40 percent devoted to microarrays. A colleague here said: ‘I don’t know what all this microarray stuff is about, but I want to start looking into it.’ I said, ‘I have heard about this through my husband.’ So, I thought I could help him and we would have a little group of statisticians talking about this stuff. I had to go and seek out the data. It was not like I was having collaborators coming to my door and saying please help me analyze this data — that wasn’t happening at Columbia yet. My sister-in-law did her residency here and I asked if she knew anybody who was working with microarrays and needed help. After speaking to a couple of people, I found someone who said he would love to have the help. He was basically analyzing the data himself, and floundering — and that was true for a lot of people who were generating microarray data. So my colleague and I started working with him on his analysis and some of his problems and we shelved the idea for a workshop.
But the workshop came up about a year later, and now, I would say is it somewhat like a pilot study. I’m seeing how it goes and if things go well, I would like to make it somewhat of an official course. We got some statisticians together and said let’s have this forum that is going to be really informal. We know nothing about microarrays, so we are just going to start from the very basics. Through a collaborator in the department of medicine we got this list of other biologists and computational biologists and other statisticians and it became this thing that now is really 30 percent statisticians, 70 percent non-statisticians. So the non-statisticians are either pure microbiologists or computational biologists. It’s this group of people where everyone is interested in the analysis of microarray data — some people knew more than others, and we started out as a journal club, every other week getting together, starting from the top. We had someone come in and give a lecture on what are microarrays, what kind of data do they generate, what are the different types of microarrays, and then I gave a little lecture on the statistical issues that people work on or the issues that come up with this kind of data. We started reviewing the papers and going through the literature together and we progressed through people presenting their own data analysis to get feedback from the group and then people started presenting their own works in progress. It is really informal.
Does the workshop help create research relationships, the kind of atmosphere that schools are spending big dollars to build—environments where different disciplines can interact?
Before the workshop, I wasn’t being approached by any collaborators who needed statistical support. So, the workshop got my name out there, as well as promoting the statistician’s role in microarray research too. And, at the time, NIH was handing back grants saying, ‘Look you need to get statistical help on this.’ I think it was a combination of all of that, and suddenly, everyone was knocking on my door and saying, ‘We need statistical support, we need help writing this grant, we need help with this data analysis.’ This definitely got the word out to the biology community, to people who are generating the data, that you do need some help, you need the perspective of a statistician here, so I think it helped in that regard.
Have you gotten any co-authorships from this effort?
One so far, and a few that are in the oven.
What is the microarray community like at Columbia now, from where you stand?
A lot of my collaborators are in the Department of Medicine and in the Department of Pediatrics I have a lot more collaborators now. There are also some new departments that have cropped up at Columbia, like the department of biomedical informatics, and the Columbia Genome Center.
Is this world becoming more organized?
I don’t really have a reference point, so it’s hard to say. It does seem as if I had to push my way into the field a little bit more than at other places. At Sloan-Kettering, where my husband works, as soon as he arrived, he had a lot of people knocking on his door and it was a little bit more systematic in that people knew to go to him.
Maybe it is because it is the clinical world, where microarrays are just becoming an important research tool?
It’s really permeating everywhere. I know of a periodontist here interested in looking at differences in gene expression between people who have a more advanced stage of the disease versus not. And, I have a collaborator who is looking at differences in gene expression between left vs right hypertrophy in children with congenital heart disease. So it’s really beginning to appear in almost every clinical field.
What do you see in the next few years?
It would be ideal in the long run to see clinicians actually use what we learn from these studies, and have some cheap version of array technology so that clinicians can use them for diagnostic purposes. That would be ideal for the future.
What are statistical problems that can be overcome?
Some of the statistical issues have to do with things like multiple testing. That’s what I am working on — trying to find genes that are differentially expressed after adjusting for multiple comparisons, so I am trying to solve that using mixture models. Right now, my short-term goal is to give exposure to the other faculty here, and the students who are going to be graduating from Columbia. It seems like there are a lot of jobs for people who have some kind of genomic background. If you look at all the job descriptions that are coming to my desk looking for graduating students, they are often looking for some exposure to this kind of data, and we don’t have any sort of course for that.
So, I would like to make [the microarray workshop] an official course and get some students to come to the classes and get some exposure to genomic data and the types of statistical issues that arise — issues of normalization, design, analysis. There are a lot of statistical issues in all of those areas. In the biostatistics department, we have a statistical genetics division. But they aren’t doing analysis for high-throughput data, they are doing classical statistical genetics, like linkage analysis and association studies. We don’t have anything in place right now where the students can get exposure to the analysis of gene-expression data, and any kind of the high-throughput data that is coming out now. Until we can design some course like that, this workshop is one venue where students can get exposure.
So basically our program is good in that it provides the basic fundamental tools that you need, but it would be nice to add that focus. Everyone recognizes that is very topical right known and it is here to stay.