Danielle Posthuma: eight hundred processors to find a gene
‘Lab techniques are making major strides, while statistical methods stay behind.’ Bioinformatics and statistical genetics are a novel and mostly unexplored field, according to Danielle Posthuma.
Danielle Posthuma studied medical anthropology and clinical and physiological psychology, but is now one of two program leaders of Integrative Analysis and Modelling. ‘I was always very interested in statistics’, she explains, ‘so for my PhD thesis on genetic causes of cognition, I devised several new statistical methods to analyse genetically informative data.’ For these purposes, she generalised and extended existing methods, Posthuma explains.
Posthuma continued her scientific career conducting research on the genetic background of cognition. She also worked on optimising statistical analyses. She witnessed how the increasing volume of data is became a huge challenge for researchers. ‘Back then, everybody performed their calculations on their own small PCs. Datasets grew and grew, and calculation times increased simultaneously. From one hour to a day, and from thereon up to a whole week and more.’
Posthuma decided to resort to SARA, the Amsterdam-based ICT centre which houses a supercomputer with multiple processors. It enabled her to do her calculations much more quickly, which was satisfactory, but, as she adds, she felt held back by the continuous need to register for available time slots. She wrote a proposal for the Netherlands Scientific Organisation to acquire a cluster computer of her own, which was approved, resulting in the purchase of a 128 node cluster computer. ‘That computer is also located at SARA, where it is connected to other computers. Together, we now have eight hundred processors.’
All programs run on UNIX, which is unfamiliar to most people working in psychology, Posthuma explains. ‘I have always liked this kind of computing and I trained myself to use this operating system.’ She also taught fellow researchers to use the supercomputer at SARA, she says. ‘The basics are not very difficult. If you know how to correctly use ten commands, you can manage the straightforward calculations.’
Her smallest dataset, Posthuma says, contains three thousand lines and 2.5 million genes, or 5 million genotypes. Posthuma is part of several international projects, one of which joins the data of eighty thousand researchers with similar datasets. All numbers multiplied, this amounts to a staggering 1.2 quadrillion (1015) records. Posthuma: ‘With six hundred processors, we can perform calculations roughly six hundred times faster than on our own pc’s. You can imagine one needs this kind of processing power for such datasets.’
Biologically meaningful mechanism
Posthuma is still searching for genetic causes of diseases and other phenotypic traits. She now wants to move beyond identifying single genes associated with diseases, she says. ‘In my experience, most genes you find may account for at most 1 or 2 percent of the variation in a disease. I want to take it one step higher.’ Her latest research delves into gene networks, which she thinks are more informative than single genes. ‘A network may consist of say, ten genes and is associated with the biologically meaningful mechanism. If one gene doesn’t function, there might not be a consequence, because other genes in the network have similar functions and can take over. If all ten genes are broken, then there might be a strong incidence for a disease or elevated score on a trait.’
Within the context of Neuroscience Campus Amsterdam, Posthuma now works on a functional gene network analysis for the traits intelligence, bipolar disorder, schizophrenia, and depression, together with VU researchers Matthijs Verhage and Peter Heutink. ‘In this project, we put in new knowledge and also developed new methods, because they did not exist yet. I believe this can be the strength of campus and our platform Integrative Analysis and Modelling. My colleague Mathisca De Gunst and I are able to answer questions arising from current research, and we can do it quickly. We also try to anticipate upcoming issues.’
With more capable bio-informaticians, she thinks, the Neuroscience Campus Amsterdam has the potential to be in the forefront of the neuroscience field. ‘We are strong in a number of research fields in biology, mathematics and genetics. Integration should be our strength.’
