A novel web portal - BrainScope - went viral in academic circles after news of it was published in Nucleic Acids Research.
This new tool, a co-development by Delft University of Technology and Leiden University Medical Centre, enables scientists to quickly and easily investigate gene activity throughout the brain. It makes patterns visible that would otherwise remain unnoticed. The only thing scientists need to run this “roadmap for the genome in the brain” is a PC with a web browser.
The BrainScope portal builds on the research efforts of the Allen Institute in Seattle, a privately funded research organisation that aims to accelerate research in bioscience. Among other things, this has led to the release of two very complex data sets (or “atlases”) of gene expression in the brain. One is an atlas of the adult human brain, based on 3,700 tissue samples derived from the brains of six deceased donors. This atlas provides access to the activity of about 20,000 human genes in the brain. The second atlas contains tissue samples from eight different stages of the developing brain, ranging from the embryonic to the adult brain. The Allen Institute is encouraging other scientists to build their research around these unique gene expression atlases, shared through www.brain-map.org.
That is exactly what the joint effort of LUMC and TU Delft was designed to achieve. "20,000 genes multiplied by 3,700 tissue samples makes an overwhelming amount of raw data. As a result, it’s hard to get a clear view of gene activity in the brain at a glance,” says pattern recognition expert Sjoerd Huisman, who is first author of the publication and a doctoral candidate in Bioinformatics at TU Delft. Thanks to elaborate processing, BrainScope offers scientists the ability to “see the data” in a very concise manner: this makes it easier to find answers to all kinds of scientific questions.
As their starting point for queries, scientists can take either locations in the brain or specific genes. In the first case, they can compare the gene activity between various tissue samples. The second starting point enables them to look into the activity of specific genes in the brain. In his pre-processing, Huisman has merged these two kinds of approaches to the data in a matrix set-up. This makes it easier to link active genes to brain regions – and the other way around.
Huisman’s “data engine” is hidden under the bonnet of an easy-to-use web portal made by LUMC scientific programmer, Baldur van Lew. The portal combines gene expression data, bioinformatics processing and imaging. It houses two different kinds of clouds as well as brain images to visualise the matter at hand. For the first, Huisman applied statistical approaches to map genes that have similar gene expression profiles and thus similar activity close together in a point cloud. The brain images, in turn, show with points where in the brain certain genes are active. The second cloud contains the tissue samples. “Together, you can regard them as two linked roadmaps,” says Huisman. “On the roadmap of the brain, every point represents a gene, and on the roadmap of the genes, every point represents a brain area.”
“The most difficult part of my task,” says Baldur van Lew, “was to translate big-data-size processing into a quick and responsive website. One of the solutions was to make optimal use of a computer’s graphics card. For the graphics, I made use of libraries that are normally used by game developers to render textures. In that way, computational power could be reserved for the core processes.”
Finding gene function
“Basically, the main task was to reduce high-dimensional data to a two-dimensional image in order to get from tissue samples to genes, and from genes to tissue samples,” says lead investigator, Professor Boudewijn Lelieveldt. “For instance, a scientist might have a hypothesis about APO-E, a gene involved in cholesterol metabolism. BrainScope not only clearly indicates where in the brain this gene is active, but can also instantly show fifty genes with similar expression in the brain. That’s new and meaningful, as genes with a similar expression profile are more likely to play a role in the same biological process. So if a gene’s activity is known, the BrainScope can lead the scientist to ‘unknown’ genes potentially involved in the same process. This could lead to an acceleration of hypothesis generation.”
With APO-E, knowledge about a gene was the starting point, using the genes data set. Alternatively, an interesting location on a brain image could be the starting point. “You can select a set of genes by just drawing a circle around them,” says Van Lew. “A database can subsequently sum up the known functions of the selected genes.” And these genes, says Lelieveldt, may be pointing at genes known to function in the immune system function. “The important thing is that, although this activity was already known in the case of a number of the listed genes, it’s new information about other genes, which have so far remained under the radar. It becomes easier to perform targeted searches for new, unknown gene functions.”
Other researchers are also interested in how gene expression measurements stemming from six different donors relate to each other. “Using BrainScope makes it possible to search for similarities in gene expression between the different donors,” says Lelieveldt. “What’s more, researchers can compare the results from these donors with their own patient data. When you investigate a disease linked to a pathway, you can select the genes involved in this pathway and see what comes out if you follow developments in the healthy brains from the Allen data set, and compare them to your own, disease-related measurements.” This could lead to new insights. The Allen data set, for instance, might point to more genes involved in the same process, thus completing the picture. This concept is being used in the LUMC translational neuroscience programme to compare people with and without brain disease, advancing our understanding of brain diseases such as migraine, autism, Duchenne’s muscular dystrophy and Alzheimer’s disease. In the same way, the data set of the developing brain can be used to compare it with the adult brain. “The bottom line,” says Lelieveldt, “is that all these comparisons can be made quickly and intuitively by using the BrainScope.”
The BrainScope project again underlines the added value of team science: gathering Medical Delta domain specialists from Delft and Leiden into a single team on bioinformatics, visualisation, image analysis and life sciencesProf.dr.ir. Boudewijn Lelieveldt, LUMC & TU Delft
“You can’t separate out the accomplishments of individual team partners. We see something similar in the Cytosplore project. There, we use the same approach and many of the same competences, but in the entirely different field of immune monitoring. Both projects are good examples of what you can achieve when you link clinical demand and technology-push. In the case of the BrainScope, it’s led to the development of a sort of sat nav for the genome in the brain,” concludes Lelieveldt.
Interview by: Leendert van der Ent