AsianScientist (Oct. 8, 2020) – The most complex structure in the known universe is a paradox. It’s right between our ears, but much of it remains beyond our understanding. The human brain contains over 86 billion nerve cells, or neurons, and some 100 trillion connections among them. Add to that neural pathways, glial cells and messenger molecules called neurotransmitters, and you’ve got a very intricate and powerful organ, one capable of probing the structure of the cosmos and laughing at silly jokes.
Scientists in Asia are now bringing the world’s most sophisticated computer technology to bear on this eternal mystery by comprehensively mapping the human brain for the first time—by as soon as 2024. This three-dimensional map, known as a connectome, could have a huge impact on everything from Alzheimer’s disease to artificial intelligence. From studies of monkey brains, to mapping the human brain and even simulating human neural activity on supercomputers, this effort could have profound consequences for both brain science and some of our fundamental questions about what it means to be human.
An Asian Connectome Consortium
In January 2020, the National University of Singapore played host to the inauguration of a bold collaborative effort to create a brain map. Synchrotron for Neuroscience—an Asia Pacific Strategic Enterprise (SYNAPSE) brings together researchers across the region to image the brain at the sub-cellular level, specifically 0.3 micrometer resolution, at a speed of one cubic millimeter per minute. Since capturing that level of detail would take years at a single facility, the work is being divided among several centers in the region while using automation and joint management of data to accelerate the project.
“Mapping the human brain at a resolution sufficient to chart the connections is a historic mission for science and technology,” said Low Chian-Ming, an associate professor at the National University of Singapore and a founding member of the initiative. “SYNAPSE will also generate technological breakthroughs in imaging, computation and artificial intelligence.”
More than 1,000 researchers are expected to participate in the undertaking, hailing from institutions such as Academia Sinica in Taiwan; Pohang Light Source-II in South Korea; Spring-8 in Japan; Singapore Synchrotron Light Source at the National University of Singapore; Shanghai Synchrotron Radiation Facility in China; and Australian Synchrotron.
The main imagining technology they will use is X-rays from synchrotrons, giant rings that can accelerate electrons close to the speed of light. When moving electrons change direction, they release high energy X-rays, which can be shunted through solid objects to image their internal structures.
“Synchrotron-based imaging is essential to achieving meaningful connectome mapping of a large animal brain because it is the only technique that combines three-dimensional sub-cellular resolution with high imaging speed to collect data for an entire brain within a reasonable time,” said Professor Hwu Yeu-Kuang, leader of the Taiwanese team at Academia Sinica.
Low and colleagues are confident of their approach because it builds on previous work under Accelerated X-ray Observation of Neurons (AXON), a set of X-ray imaging technologies perfected by an international scientific collaboration which is now the core of the SYNAPSE project. This technique demonstrated the speed and resolution possible on the fruit fly brain and is on its way to being tested on larger animal brains. The researchers also managed to partially map a mouse brain, according to a study in the Chinese Journal of Physics.
“Our first data set for the fruit fly brain, published in a study led by Professor Chiang Ann-Shyn at the National Tsing Hua University, proved that even results for small brains and partial information can revolutionize neurobiology research,” Hwu added.
SYNAPSE will incorporate petascale high performance computing facilities at the National Supercomputing Centre (NSCC) Singapore, which is a sponsor of the project. SYNAPSE is expected to produce a huge volume of data that will be in excess of one billion gigabytes, or one exabyte, which is the equivalent of a million one-terabyte drives. The NSCC will serve as a data hub for the processed 3D mapping data, linking processing facilities like Fugaku at the RIKEN Center for Computational Science, Taiwania at Taiwan’s National Center for High-Performance Computing, and systems at the Daegu Gyeongbuk Institute of Science and Technology, South Korea, via a high-speed 100 Gbps network run by the Singapore Advanced Research and Education Network (SingAREN).
“We specifically tested the parallel involvement of different synchrotrons and their calibration, so that data can be simultaneously analyzed and processed with uniform high performance computing platforms to be shared among partners,” said Low. “Our task is to provide neuroscientists with something like a Google Map of a complete functional brain, where they can view different levels of information as they zoom in, balancing the need for detailed information along with retrieval speed. It will also serve as a reference to chart the functional readings obtained through fMRI, electrophysiology, EEG and other techniques, building up to a structure-function brain map.”
Other imaging techniques that the SYNAPSE researchers are using include infrared spectromicroscopy, super-resolution visible light 3D microscopy and cryoelectron tomography. The information on specific brain regions obtained from these complementary techniques would then be used to annotate the brain map with a large variety of functional and high-resolution information related to the neural network.
In being able to shed this much light on the structures of the brain, SYNAPSE may give rise to something that’s long been a staple of science fiction: simulating a human brain on a computer. Since it will be based on neural network data from real brain maps, it’s expected to generate more understanding about brain functions, complementing initiatives such as the BRAIN Initiative in the US and the Human Brain Project in Europe, according to Low.
“No individual country would have the resources to finish this project in a reasonable amount of time. Parallel data collection is essential and can only be realized by an international project,” Hwu said. “The same is true for processing and storing the huge amount of data, and for further developments of the imaging and data processing techniques.”
In the same spirit, the results will be shared by all partners and open to everyone, with the help of cloud computing and high performance computing.
“As the raw data size is in exabytes, each country partner will store the raw data but deposit the reconstructed core data at NSCC Singapore,” Low explained.
In the first phase that is expected to be completed by 2024, SYNAPSE aims to map one human brain.
“However, the main goal of our collaboration is to understand how the human brain functions from the complete mapping of the neural network. This could lead to effective therapies for brain diseases, a global problem with a huge social impact,” Low said. “This mission may take more than a decade.”
China’s long play
SYNAPSE follows a number of other large-scale brain computing endeavors in East Asia. One is the China Brain Project (CBP; also known as the China Brain Initiative), an enormous initiative stretching to 2030 with multiple aims: expanding basic research on cognitive functions; applied research into the diagnosis and treatment of brain disorders; and research into computing methods inspired by the human brain.
The CBP is intended to rival similar efforts in the West, and promoters have said that compared to other initiatives, the CBP is prioritizing brain disorders and artificial intelligence as its near-term goals. Part of that is due to the fact that China, with its enormous and aging population, is saddled by large numbers of people with neurological and cognitive disorders.
Professor Poo Mu-ming, leader of the CBP and director of the Institute of Neuroscience of the Chinese Academy of Sciences, said the CBP has not officially started yet, and referred Supercomputing Asia to a 2016 article he co-authored in the journal Neuron which stated that the CBP is part of a 15-year plan.
“The China Brain Project covers both basic research on neural mechanisms underlying cognition and translational research for the diagnosis and intervention of brain diseases as well as for brain-inspired intelligence technology,” Poo and colleagues wrote.
While there have been few details of this grand national undertaking, a number of new research centers have opened. The Chinese Institute for Brain Research, Beijing, was inaugurated in 2018 as a main pillar of the CBP. It has been engaged in recruiting 50 internationally selected principal investigators and more than 1,000 other researchers as well as building 11 core facilities with some 30,000 animal cages. A sister center, the Shanghai Research Center for Brain Science and Brain-Inspired Intelligence, was also launched in 2018.
The CBP’s emphasis on animal research comes after a team led by Poo produced two genetically identical long-tailed macaques, a world first for primates, using somatic cell nuclear transfer, the technique that created Dolly the sheep. The study, reported in the journal Cell, raised hopes that the cloning know-how could be used with gene-editing techniques such as CRISPR-Cas9 to recreate monkey models of human brain disorders such as Parkinson’s disease.
“This paper really marks the beginning of a new era for biomedical research,” Professor Xiong Zhi-Qi, a neuroscientist at the Chinese Academy of Sciences Institute of Neuroscience, was quoted as saying by Nature News.
Similar work has broadened the field: Chinese scientists have even created transgenic macaques with extra copies of a human gene that might have a role in human intelligence. They reported that the enhanced monkeys outperformed their peers on a memory test. As if alluding to science-fiction scenarios from Planet of the Apes, Western scientists have questioned the ethics behind tinkering with monkey genes and intelligence.
Supercomputing brains in Japan
Unlike the CBP which focuses on macaques, Japan’s Brain Mapping by Integrated Neurotechnologies for Disease Studies (Brain/MINDS) looks at another non-human primate: marmosets. Brain/MINDS was launched in 2014 and is now in its second phase. With the state-backed RIKEN Center for Brain Science (RIKEN CBS) playing a coordinating role, Brain/MINDS brings together universities and research institutes from throughout the nation.
The project’s goal is to elucidate the neural circuits underlying higher brain functions using unique experimental models. A particular focus is neural circuits that are responsible for neurological and psychiatric disorders. As member researchers wrote in Philosophical Transactions of the Royal Society B: Biological Sciences, the objectives fall into three main areas: functional mapping of the marmoset brain, developing innovative neurotechnologies for brain mapping and actual brain mapping, along with clinical research.
Recent scientific reports from the initiative include the role of auditory signal processing in schizophrenia, neuroinflammation in mouse models of Alzheimer’s disease, and a comparative study of how common marmosets and Japanese macaques react to human behavior. RIKEN’s Professor Hideyuki Okano, dean of Keio University’s Graduate School of Medicine, has been producing genetically modified marmosets to learn more about Rett Syndrome and Parkinson’s disease.
“The marmoset brain is essentially a simpler primate brain to study and its common features to other primates make it easier to analyze,” said Professor Alexander Woodward, leader of the Connectome Analysis Unit at RIKEN CBS, where the goal is to chart and analyze brain connections. “The developmental period is much shorter than that of humans. Furthermore, the brain does not have the gyrification (wrinkled appearance) that larger primates or the human brain have, making it easier to carry out certain invasive experiments.”
A key part of these experiments is tracer injections. As Woodward explains, a viral tracer injected into the brain infects neurons, causing them to express a fluorescent protein that travels down the axon to the terminal sites. By later examining the brain ex vivo, researchers can see the connections that neurons at the injection site make with neurons in other brain regions.
Brain/MINDS scientists are also using MRI technology known as diffusion weighted imaging (dMRI) to obtain structural data. This data can provide insights into the overall connectivity pathways and the structural pathologies where there is damage to the brain connections. Members of Woodward’s team also developed a 3D marmoset brain atlas which describes all of the brain regions and this is being used to calculate and summarize the brain connectivity patterns from the data.
Pushing the frontier with Fugaku
Another major RIKEN project related to the Brain/MINDS endeavor is the construction of a next-generation supercomputer named Fugaku (see our infographic). An alternate name for Mount Fuji, Japan’s tallest peak, Fugaku is designed to have an ARM architecture and some 150,000 CPUs. It would be at least 40 to 100 times more powerful than its predecessor, the K computer, which TOP500 ranked as the world’s fastest supercomputer in 2011. Under construction by Fujitsu, Fugaku could become the world’s first computer with exaflop speed, or 1018 floating-point operations per second, if it enters full operations around 2021 as planned and doesn’t lose out to rivals in China, the US and the European Union.
Fugaku will be used for everything from drug discovery to simulating earthquakes and tsunamis. Neuroscientists in Japan, though, are eager to use it for brain simulations. They believe Fugaku will be the first computer with sufficient resources to simulate a human brain—it will recreate the same number of neurons as the human brain as well as the connections between them. The feat will be astonishing given that only seven years ago, it took the K computer 40 minutes, using 1.73 billion virtual nerve cells connected by 10.4 trillion synapses, to simulate only one second of neuronal activity in a human brain. Japanese researchers have already begun simulations on part of Fugaku and, with the full system, they plan to test the computational performance of a brain model consisting of the cortex, cerebellum, thalamus and basal ganglia.
“Our goal is to realize whole-brain simulation using the Fugaku computer for an understanding of the mechanisms of brain function and disease,” said Dr. Jun Igarashi of the High Performance Artificial Intelligence Systems Research Team at RIKEN’s Center for Computational Science. “As a first step, we plan to perform whole-brain simulation of mouse, marmoset and human, to understand the interactions between brain areas in processing information for movement and behavior.”
Using machines like Fugaku, scientists may be able to better understand the brain’s fundamental information-processing ability as well as activity, such as how movement is coordinated over a span of seconds to minutes. But will researchers actually be able to simulate a mind on a computer? That remains to be seen.
“No one knows whether whole-brain simulation can reproduce human intelligence until it is realized,” Igarashi said. “The current problem is that there is no consensus on which levels of description of the neuron are sufficient for reproducing intelligence.”
Bolstered by grand initiatives like these projects across Asia, scientists hope to one day gain a complete mechanistic understanding of the universe’s most complex object. But that doesn’t mean our brains can be reduced to biological computers. If anything, it’s the part of us that can’t be simulated—the ghost in the machine—that makes us human.
This article was first published in the print version of Supercomputing Asia, July 2020.
Click here to subscribe to Asian Scientist Magazine in print.
Copyright: Asian Scientist Magazine.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.