Hat trick for Alzheimer’s grand challenge

Biomarker imaging output displayed on an interface for neuroscientists to analyze. Image courtesy Montreal Neurological Institute, McGill University, Canada.

A hat trick has occurred in scientific computing. Scientists are using three e-infrastructures in three countries, with three separate technologies, to identify biological signatures of neurodegenerative diseases in human brains. We previously reported on the outGRID project and its ‘grand challenge’ to help neuroscience laboratories around the world work on large-scale experiments. Since then, the challenge has come to be known as the LINked Neuroscientific Grand chAllenge (LINGA), which involves more than 2,000 CPU cores working in parallel around the world.

Brain-scan imaging data are hosted in three separate Alzheimer’s disease neuro-imaginge-infrastructures. Each is based on different technologies. neuGRID is grid based, LONI (Laboratory of Neuro Imaging) is cloud based and CBRAIN runs on HPC. This computing pipeline helps neuroscientists in different locations around the world work in real time to select, process and analyze biomarkers of Alzheimer’s disease.

The output of these analyses are 3D brain representations (as shown below), which are then aggregated into and statistically analyzed by the European Grid Infrastructure (EGI). Then, neuroscientists compare selected factors such as age and sex of individuals to find potential new correlations. It is hoped new insights into Alzheimer’s will be found and technical interworking between the participating infrastructures will improve.

This working collaboration has an overall coordinating body. David Manset, technical coordinator of outGRID showed a working demonstration of the LINGA infrastructure at the EGI Technical Forum in Lyon, France, last week. “SHIWA [Sharing Interoperable Workflows for large-scale scientific simulations on Available DCIs], is the federating platform on top of this network. It is an EC FP7 funded project that enables this distributed infrastructure to connect Canadian, European and US institutions simultaneously, thanks to a unique simulation platform. If done on a single infrastructure, this would take years, but parallelizing the work has speeded things up. Four separate interfaces can be connected into one harmonized environment so new markers can be developed and tested on large datasets,” said Manset.

The LINGA challenge was awarded best live demonstration at the EGI Technical Forum. It was executed by six different research groups and projects: LONI at UCLA in the USA, CBRAIN at McGill in Canada, neuGRID/outGRID in Europe, SHIWA in Europe and EGI and EMI in Europe.

In August 2011, Manset and his colleagues published a paper in Nature Neurology about Virtual imaging laboratories for marker discovery in neurodegenerative diseases.