At A Glance
- 2001-present — Bioinformatics research associate; Scottish Centre for Genomic Technology and Informatics.
- 1997-2001: Medical Statistics and Database development; Public Health Sciences, University of Edinburgh.
- 1994-1997: Degree MedDok (“Medizinische Dokumentation”): Justus-Liebig University, Giessen, Germany
Thorsten Forster, 29, is the lead author of the article “Experiments using microarray technology: limitations and standard operating procedures” published in the August edition of The Journal of Endocrinology. The bioinformatics research associate at the Scottish Centre for Genomic Technology and Informatics in Edinburgh, UK, and his collaborators have created a system of standard operating procedures for biology newcomers who are interested in engaging in microarray analysis.
BioArray News caught up with Forster recently to talk about the article and about how the standard procedures were created.
Who was this article written for?
We were addressing the bench biologists, people just starting out in the field. The problem seemed to be that lots of people work on normalization analysis methods, experiment design, but there wasn’t a workflow to help when they wanted to start out. I know that, because we collaborate with a lot of them. What we [needed] was an SOP. So, I sat down and tried to devise one.
This entire field is more than a bit of [a] maze. What we are trying to supply here is a bit of bread crumbs. So that with the maze of microtechnology, you don’t have to be an expert, you don’t need to know it to find your way through it, you just have enough bread crumbs to get to the end of the maze with some reasonable results.
What did you learn in the process?
I learned what bits are confusing to the user, what bits need to be rehighlighted because they haven’t been done to that extent in the world of biology, as far as I know, anyway. There is experiment design, which hasn’t been a big issue in biology because you had three animals here, three animals there, and you did a bar chart for those values. Now, with bioarrays in general, the issues in analysis have increased manifold — so that had to be included and simplified in some way. So, in going through this, and trying to figure [an SOP], we learned how many options there are, how many if-then-elses there are out there.
How long did it take to create?
I’ve been working here for two years, and I started working on this in week two. It took a long time to go through the experiment design, and the scan selection, because there were a number of scanners we and a number that others are working on, and we needed this to be fairly general, and work for everybody — and make sense as well. All of this took maybe a year and a half, from beginning to end.
In the referee process, what were the comments that you received?
They were mainly in positive notes that said, ‘Yes, we need some.’ There were, of course, the usual notes: ‘Not enough detail.’ To which we could only reply, ‘yes, but that is not the point though. If we wanted more detail, we would have to send people off to experts.’ What we wanted was something that works for most of the people, most of the time. Apart from that, it was only minor things that were requested to be changed, bits of formula too complicated, or too easy, that sort of thing.
Microarray analysis is quite a dynamic field, so it’s hard to take a snapshot isn’t it?
I fully realize that what we have taken is already out of date. But, you have to start taking snapshots at some stage, otherwise, no one will do it. We got [Journal of Endocrinology editors] interested in doing a SOP because when researchers were starting out in microarrays, they looked at what was out there — all the papers, all the websites — and decided this is way too complicated. So, you need an introduction, a guideline.
What kind of microarray technology is used at The Scottish Centre for Genome Technology?
We have a number of printers. We use MWG Biorobotics, and the Affymetrix platform, which we run as clients — we buy the chips, we have the scanner. We work as a collaborative center. We are not really here to provide a service for everybody who wants to do microarrays. We collaborate with people. We train them in the use of the instrumentation. We might design and print the arrays for them and with them. We collaborate and maybe we get some research out of this and enhance our methods and procedures and protocols. Microarrays are only one of our areas. It’s functional genomics — with microarrays at one end and the transcriptome at the other end, and this needs to go further to clinical data. So we are collaborating with those people as well to get a greater pool of data and integrate all in the end.
What’s your throughput in terms of arrays printed a year?
We print a few hundred, to a few thousand. We do continuous testing, this scanner versus that scanner. The more you do, the more experience you get in the problems you come across.
Have you considered other platforms?
We are also using the Agilent platform, and I can tell you that because I saw someone using the chip. I haven’t come across it in terms of data analysis, yet. We are starting out testing the system and applying it to various research problems before it comes my way in terms of data analysis. I mainly deal with the experiment design and the subsequent analysis
What is your view of the microarrays in Europe?
You are of that generation of young biologists.
Bioinformatics in the UK is very young. Only in the last few years have there been courses where people learn that. It used to be that biologist learned how to use algorithms, and maybe how to design them. Microarrays and proteomics have totally changed that. So, now we need statisticians, and hard-core computer scientists to do the programming work. They now become bioinformaticians although they never had any formal education as bioinformaticians, and they are learning the biology.
What would you like to see improved in the microarray technology?
I would like [it] to be a totally reliable tool that you can use every day by just having a sample and a one- or two-day routine to figure out what it does genetically speaking. And then you use that data combined with clinical data and proteome data to get to the bigger picture. Transcriptomics doesn’t mean proteomics and proteomics doesn’t mean phenomics, or what happens at the level of the individuals. So, microarrays become a tool for learning more about the rest, and it should be that reliable.
I’m not a biologist, but that is just my term. I try my best to avoid these words, because I get annoyed with them. I think the latest word I came across is integromics, integrating everything with everything else. Heard it in a conference talk somewhere.
Is science being rigorous enough with microarray results, the reproducibility, and the accuracy?
We are a few years away from that. We don’t have the resources in academics to produce hundreds of thousands of arrays on a regular basis. So, we are aware of what we are producing is a data reduction technique where we can boil down all of the stuff off the genomes to 20, 30, 40 interesting ones to follow up later, which is one of the most important things to do. Most experiments will never achieve the total statistical significance that is required because the resources aren’t there.
What can the scientific community do to add integrity to this platform?
Follow the MIAME folks. We try to push that as hard as we can internally, but we are probably falling short about using some of the terms. The principle that you should annotate what you are doing, and document what you are doing, that applies. And even small mistakes can lead to great problems in the end in data analysis.
There are protocols that need to be followed to the letter, and that requires a very conscientious person. Biologists are usually very independent people, they like to do their own thing. They need to be made very aware that if you do this, the consequence will be that, in terms of data processing. If you don’t shake it vigorously seven times, and you only do it twice, then you might have this result in the end from the data analysis. The awareness of that is spreading. The idea that, hey, there are statisticians and a bioinformaticians, and they can help me. Bioinformatics was developed to have a bridge from biology and computer scientists. I wouldn’t suggest that the bench biologists all need to become good bioinformaticians to do microarray work. I’m firmly convinced they should be biologists and know what is important, and they should know who to talk to make things work.