Skip to main content

Human Brain Project Specific Grant Agreement 2

Periodic Reporting for period 1 - HBP SGA2 (Human Brain Project Specific Grant Agreement 2)

Okres sprawozdawczy: 2018-04-01 do 2020-03-31

Understanding human brain organisation at all relevant levels is a big challenge, yet this must be done to improve treatment of brain disorders, create new computing technologies and provide insight into our humanity. The HBP’s unique strategy develops cutting-edge ICT to gather, integrate and analyse brain data, understand the brain, healthy or diseased, and emulate its unique computational capabilities. By sharing our tools and data, we aim to catalyse collaboration and accelerate progress in these fields.
The HBP’s high-level objectives for SGA2 were to:
1) Establish, operate and disseminate the EBRAINS digital research infrastructure, driven by scientific use cases and based on the existing individual Platforms.
2) Establish the gathering, organisation and dissemination of neuroscience and medical data as the core of EBRAINS, bridge different levels of the multi-level brain organisation, address inter-subject variability to understand changes over time, between sexes and in diseases, and create from these a scientific impact for neuroscience and brain related medicine.
3) Establish the workflow from data to models, built and simulated with external community involvement using EBRAINS, with results fed back into empirical neuroscience and novel computing architectures.
4) Establish the connection between the different levels of the multi-scale models of brain organisation, brain changes over time and multi-scale theories for healthy and diseased brains.
5) Develop neuromorphic computing, high-performance computing and neurorobotics into a pioneering approach to biologically inspired AI, with an emphasis on continuous learning and cognitive computing, and enable extreme-scale computing for neuroscience simulation and data science applications.
6) Establish responsible innovation, open data and open science as guiding principles of HBP research, which beneficially affects society as a whole.
The work was organised around 12 Subprojects (SPs) and horizontal Co-Design activities. The HBP has created a large collection of curated multi-level, multi-species data, tools and models, from single neurons to whole brain, which makes EBRAINS (https://ebrains.eu/) a unique community resource. Cross-disciplinary, multi-modal approaches, across species and methodologies, have provided unprecedented insight into brain organisation at different levels (the SPs indicated below are the main contributors to the work mentioned, but are not the only ones).
The HBP Atlas provides access to ultra-high resolution anatomical data in the BigBrain, plus proof-of-concept using iEEG data via Hibop, and a link to The Virtual Brain simulation/modelling support (SPs 2 & 5). This allowed fusing computational modelling, deep learning, experimentation and robotics to understand how the brain coordinates visually-guided actions (SPs 2, 3 & 4), resulting in a new closed-loop architecture for robotic systems (SP10). Developed to support drug discovery, a first automated GPCR-tailored protocol connecting molecular and subcellular levels enabled prediction of ligand binding at subcellular level (SPs 2 & 8). Integration of molecular properties was also shown in large-scale mouse brain cellular simulations, run on HPC resources provided by Fenix (SPs 1, 6 & 7). A framework was developed for assessing consciousness and brain complexity in human, rodent and in silico brains (SP3).
EBRAINS made selected services available to external users. The Neuroinformatics Platform improved input into the Knowledge Graph; its data and models content tripled in the last 12 months (SP5). The brain atlases were updated to cover more regions, template spaces and modalities (SPs 1, 2 & 5). A new Collaboratory was introduced (SP5). Brain Simulation Platform use increased and new tools were added for modelling at different biological scales (SP6). The HPAC Platform has developed data federation and data-intensive computing services, such as the Fenix Data Transfer Service and Central Data Location Service and improved simulation tools’ readiness for HPC systems (SP7). The Medical Informatics Platform is deployed in 30 hospitals, with 20,000 datasets in dementia, mental health, epilepsy and traumatic brain injury – a scale offering statistical significance (SP8). The Neuromorphic Computing Platform (SP9) unveiled a one million-core SpiNNaker system (the largest-ever neuromorphic platform) and supported cross-SP work on learning and motor control loop experiments, plus application of structural plasticity to networks processing auditory and visual stimuli. The Neurorobotics Platform (SP10) improved its usability and interoperability with other EBRAINS services, including HPC deployment, and supported spinal neurorehabilitation work, with publications in Nature and Nature Neuroscience.
In its next phase (SGA3), the HBP will focus on key scientific showcases and key elements of EBRAINS, for which an EBRAINS AISBL has been created. Nine Calls for Expression of Interest, to bring in new Partners, and Infrastructure Voucher Calls to support Partnering Projects, helped prepare for SGA3. The HBP developed a proposal for EBRAINS to join the European Strategy Forum on Research Infrastructures (ESFRI) Roadmap, to ensure its availability after the end of the FET Flagship.
The Virtual Epileptic Patient started its first clinical trial, to improve surgical outcomes for epilepsy patients (SPs 4, 5 & 8). The Human Brain Atlas offers an unprecedented “window on the brain”, able to “place” scientific findings in their topographical context (SPs 2 & 5). Structural complexity of neurons is better understood, notably for pyramidal and cerebellar granular cells (SPs 1 & 2). Mouse brain work led to a model of the striatum (SP1), which is key for motor control, and the human auditory cortex is now understood in much finer detail (SP2). Mathematical models of wave-like activity and assembly projections improve our ability to create generic models of spiking neuron networks (SP4). HBP researchers filed three patents for improving drug target identification (SPs 2 & 8).
EBRAINS provides unprecedented access to services for brain research and brain-inspired sciences, helping scientists to collect, analyse, share and integrate brain data, and apply them to models and simulations. Interoperability between platforms is key; connecting the Knowledge Graph (SP5), Brain Simulation Platform (SP6) and Neurorobotics Platform (SP10) allows brain models to be imported in SONATA format for use in robotic experiments. Exploitation of neuromorphic computing systems (SP9) in neurorobotics is also promising. Integration of The Virtual Brain into the HBP boosted infrastructure capabilities (SP5). HBP Partnering Projects increased from seven at the end of SGA1 to 40.
The Education Programme (SP11) showed how to apply HPAC to advance neuroscience and, with Ethics and Society experts (SP12), provided training in responsible research and innovation. The HBP also published influential journal articles such as “The Scientific Case for Brain Simulations” (Neuron) and the HBP’s “Synergy between neuroscience, computing, informatics, and brain-inspired technologies” (PLOS Biology). SP6 created an innovative outreach tool in its “Live Papers”, on line resources which combine texts with datasets and software used to obtain published results.