Markus Diesmann, Ph.D.

Future brain models will not be created by individuals but by increasingly larger teams of researchers because of the complexity of the undertaking. At the outset of the Human Brain Project (HBP) the reproducibility of published network models and data analysis in computational neuroscience was limited. Thus, the community was ill-prepared for the new era, both technologically and sociologically. This lecture introduces, on the example of a multi-scale network model of one hemisphere of macaque vision-related cortex, the progress made in technology and in transforming the way computational neuroscience is done.
The cortical network exhibits organization on multiple levels but an integrated view is missing. In particular, it has been known for a long time that cortical architecture, the area-specific cellular and laminar composition of the network, is related to the connectivity between areas, forming a hierarchical and recurrent network at the brain scale. Our recent study [1] integrates data on cortical architecture and axonal tracing data into a multi-scale framework describing one hemisphere of macaque vision-related cortex at cellular resolution. Simulations [2] confirm a realistic activity regime. At a sufficiently large coupling between the areas, spike patterns, the distribution of spike rates, and the power spectrum of the activity are compatible with in-vivo resting-state data. Furthermore, the matrix of correlations between the activities of areas is more similar to the experimentally measured functional connectivity of resting-state fMRI than the anatomical matrix. Thus, the model relates the single-neuron level to the level of common brain imaging data. The correspondence on multiple spatial scales is achieved in a metastable state exhibiting time scales much larger than any time constant of the system.
In the initial phase of the HBP we published as open-source code the formal executable description of a model of a cubic millimeter of cortical tissue [3] containing all its roughly 100,000 neurons and 300 million synapses between them. At this scale neurons reach their natural number of 10,000 synapses. The model is based on the integration of knowledge of some 50 experimental studies and reproduces a number of prominent features of neuronal activity, like the distribution of spike rates across cortical layers and the comparatively high activity of inhibitory neurons. This model of the cortical microcircuit found resonance in the community and. the neuroscientific results were quickly reproduced. The model is now in use as a building block for more advanced models, as a testbed for studies on brain function, as well as for the validation of theoretical work and neuromorphic systems. To date 17 peer-reviewed studies use the model and 71 cite the original work.
A hemisphere of the macaque vision-related cortex contains about 1 billion neurons, the limit of what supercomputers can simulate today but too costly for routine laboratory work. Therefore, our model represents each of the 32 areas by just one microcircuit adapted to the area-specific cellular and laminar composition, reducing network size to a few million neurons.
Further increasing model size, for example to faithfully represent the vastly different relative extents of cortical areas, requires progress in simulation technology. The HBP therefore is a driver of the development of the next generation of supercomputers, so-called exascale systems. The most recent code [4] developed in collaboration with the post-K computer project makes the memory consumption of the individual compute nodes of a supercomputer fully independent of total network size.
Reaching out to the brain scale requires a new approach to model development, not because of the increased model size in terms of neurons and synapses, but because of the amount and heterogeneity of the data that need to be aggregated. Therefore, we adopted methods of software development like version control, collaborative development, and code review to our needs using the GitHub platform. When working on making the model accessible for our colleagues we noticed that the executable model description is not sufficient. The experimental data entering the model span multiple scales and come from different sources. Algorithms are required to collocate the data and derive the final model parameters. In several instances data are only partially available, such that predictive connectomics is needed to formulate quantitative hypotheses bridging the gaps. As a consequence, researchers can only add new data to the model or modify assumptions if they have access to the construction process. Therefore, the workflow of data integration also needs to be documented in an executable format. Borrowing techniques from computer science and systems biology [5] we present a digitized workflow of model construction reproducing all figures of our respective publications.
The open model repository is https://inm-6.github.io/multi-area-model and a tutorial video is located at https://youtu.be/YsH3BcyZBcU.
This research used resources of the K computer at the RIKEN AICS. Supported by the project Exploratory Challenge on Post-K Computer (Understanding the neural mechanisms of thoughts and its applications to AI) of MEXT. Use of the JUQUEEN supercomputer in Jülich was made possible by VSR computation time grant JINB33. The development of NEST is guided by the NEST Initiative. Partial funding comes from the Human Brain Project through European Union 7th Framework Program grant 604102 (Ramp Up Phase) and EU Horizon 2020 research and innovation program grants 720270 (SGA1) and 785907 (SGA2).

1. Schmidt, M. et al., 2018. A multi-scale layer-resolved spiking network model of resting-state dynamics in macaque visual cortical areas. PLoS Computational Biology, 14(10), pp.1–38.
2. Schmidt, M. et al., 2018. Multi-scale account of the network structure of macaque visual cortex. Brain Structure and Function, 223(3), pp.1409–1435.
3. Potjans, T.C. & Diesmann, M., 2014. The cell-type specific cortical microcircuit: Relating structure and activity in a full-scale spiking network model. Cerebral Cortex, 24(3), pp.785–806.
4. Jordan, J. et al., 2018. Extremely Scalable Spiking Neuronal Network Simulation Code: From Laptops to Exascale Computers. Frontiers in Neuroinformatics, 12(2).
5. Koster, J. & Rahmann, S., 2012. Snakemake--a scalable bioinformatics workflow engine. Bioinformatics, 28(19), pp.2520–2522.

Kenneth Harris, Ph.D.

The brain processes information through the simultaneous activity of large populations of neurons. Modern experimental techniques make it possible to record the activity of very many of these cells simultaneously. I will describe two projects that employed these technologies to study how neurons in visual cortex, and across the brain, process sensory information and guide behavior.
The first study employed 2-photon calcium imaging to record the responses of >10,000 neurons in the visual cortex of awake mice, to thousands of natural images. The recorded population code was high-dimensional, with the variance of its dimensions following a power law. This power law did not reflect the statistics of natural images, as it persisted even when presenting spatially whitened stimuli. This dataset was large enough to allow us to consider a limit that the number of recorded neurons and stimuli tends to infinity: by considering neurons and stimuli to have been randomly sampled from a probability distribution, we are able experimentally characterize the geometry of the full population code, to the full distribution of stimuli. A mathematical analysis showed that neural population vectors lying in a set of fractal dimension d must have variances bounded by a power law of exponent 1+2/d. By recording responses to stimulus ensembles of varying dimension, we showed this bound is saturated. We conclude that the manifold of neural population responses is as rough as is possible without exhibiting fractal geometry.
The second study investigated how mice perform a visually-guided perceptual decision task. Subjects were trained to give one of three responses (choose left, right, or neither) depending on the relative contrast of two simultaneously presented visual stimuli. Widefield imaging of dorsal cortex revealed that after stimulus presentation, activity progressed from primary visual cortex to secondary visual areas, secondary motor cortex, and finally primary motor cortex. Optogenetic inactivation of visual cortex and secondary motor cortex impaired task performance, with the critical time-window for inactivation being earlier in visual cortex. Inactivating primary motor cortex, however, did not impair performance.
We used Neuropixels electrodes to record the activity of >20,000 neurons across the brain during task performance. These arrays span ~4 mm of tissue and thus record simultaneously across diverse brain regions, including sensory, parietal, frontal, and motor isocortex; thalamic nuclei; hippocampus; striatum; superior colliculus; and multiple midbrain structures.
Neurons responding to visual stimuli or predicting decisions were localized to specific brain regions, but neurons with correlates of ongoing movement or recent reward were widespread. Visually responsive neurons were found in superficial superior colliculus, visual cortex, and striatum. Neurons that predicted the animal’s choice substantially prior to movement (~100ms) were found in deep superior colliculus and the mesencephalic reticular formation. However, activity concurrent with action execution and following reward delivery were observed in nearly every region we recorded.
We suggest that when animals perform this task, visual information flows through visual and secondary motor cortices and striatum, to the midbrain where a behavioral choice is selected. By contrast, corollary information about ongoing movements and rewards is represented globally including in primary motor cortex, but this activity is not required for task execution.

1. Stringer, C. et al., 2018. Spontaneous behaviors drive multidimensional, brain-wide population activity. bioRxiv, p.306019.
2. Stringer, C. et al., 2018. High-dimensional geometry of population responses in visual cortex. bioRxiv, p.374090.
3. Steinmetz, N.A. et al., 2018. Distributed correlates of visually-guided behavior across the mouse brain. bioRxiv, p.474437.

Michael Hawrylycz, Ph.D.

We will survey the main new resources from the Allen Institute for Brain Science, including the Cell Types Database, The Brain Observatory, and Mouse Connectivity Atlas. The presentation will involve demonstrations of the brain cell database which contains a survey of biological features derived from single cell data, from both human and mouse. This database is creating a census of cells in the mammalian brain and contains electrophysiological, morphological, and transcriptomic data measured from individual cells, as well as models simulating cell activity.
The Institute also has an active research program focused on modeling the activity and behavior of the mammalian brain involving use of brain-wide, circuit-level and cell-level biophysical models, and modeling tools. In terms of dynamics, The Allen Brain Observatory presents the first standardized in vivo survey of physiological activity in the mouse visual cortex, featuring representations of visually evoked calcium responses from GCaMP6-expressing neurons in selected cortical layers, visual areas and Cre lines. The final resource we will discuss is a high-resolution map of neural connections in the mouse brain, built on an array of transgenic mice genetically engineered to target specific cell types. I will highlight and demonstrate the various tools and ways of data access from each of the these major efforts. While the main talk will be a high level survey there will be opportunity for further discussion and more detailed questions.

1. Bakken, T.E. et al., 2016. A comprehensive transcriptional map of primate brain development. Nature, 535(7612), pp.367–375.
2. Hawrylycz, M. et al., 2016. Inferring cortical function in the mouse visual system through large-scale systems neuroscience. Proceedings of the National Academy of Sciences, 113(27), pp.7337–7344.
3. Puchalski, R.B. et al., 2018. An anatomic transcriptional atlas of human glioblastoma. Science, 360(6389), pp.660–663.

Valentin Nägerl, Ph.D.

The human brain is the most enigmatic structure in the known universe - at least that's what our mind is telling us! Admittedly, its physical design is stunningly convoluted and miniaturized, with some 80 billion neurons, half a million kilometers worth of axon cables and 1015 synapses packed into a cubicle less than 1.5L in size, forming an epically powerful and versatile organ that thrives on abstract information like piano concertos, laser manuals or Haiku poems.
The advent of super-resolution microscopy has created unprecedented opportunities for multi-scale analyses of this anatomical complexity and its dynamics in a live setting using the mouse brain as a model system. I will present our recent methodological advances based on super-resolution microscopy to 1) reveal morphological mechanisms of activity-dependent pre- and postsynaptic plasticity, 2) visualize the morphological and molecular organization of synapses, 3) analyze the turnover of synapses in the hippocampus in vivo and 4) image the extracellular space of brain tissue.

1. Tønnesen, J. et al., 2014. Spine neck plasticity regulates compartmentalization of synapses. Nature Neuroscience, 17(5), pp.678–685.
2. Chéreau, R. et al., 2017. Superresolution imaging reveals activity-dependent plasticity of axon morphology linked to changes in action potential conduction velocity. Proceedings of the National Academy of Sciences, 114(6), pp.1401–1406.
3. Arizono, M. et al., 2018. Structural Basis of Astrocytic Ca 2 Signals at Tripartite Synapses. SSRN Electronic Journal.
4. Inavalli V.G.G.K. et al. A super-resolution platform for correlative single molecule imaging and STED microscopy (in revision at Nature Methods).
5. Pfeiffer, T. et al., 2018. Chronic 2P-STED imaging reveals high turnover of dendritic spines in the hippocampus in vivo. eLife, 7, pp.1–17.
6. Tønnesen, J., Inavalli, V.V.G.K. & Nägerl, U.V., 2018. Super-Resolution Imaging of the Extracellular Space in Living Brain Tissue. Cell, 172(5), p.1108–1111.e15.

Atsushi Nambu, M.D, Ph.D.

The basal ganglia are a group of subcortical nuclei composed of striatum, subthalamic nucleus (STN), globus pallidus (GP), and substantia nigra (SN). They control voluntary movements through the thalamus and cortex. The striatum and STN are the input structures of the basal ganglia, while the internal segment of the GP (GPi) and SN pars reticulata (SNr) are the output nuclei. The following three pathways connect the input and output stations and modulate GPi/SNr activity: the cortico-STN-GPi/SNr hyperdirect, cortico-striato-GPi/SNr direct, and cortico-striato-external GP (GPe)-STN-GPi/SNr indirect pathways. A signal through the direct pathway inhibits a specific population of GPi/SNr neurons, resulting in disinhibition of the thalamus and cortex and an exclusive release of a selected motor program at a selected timing. On the other hand, signals through the hyperdirect and indirect pathways excite the surrounding wide areas of the GPi/SNr, resulting in inhibition of the thalamus and cortex, and suppression of other competing motor programs.
Malfunctions of the basal ganglia result in severe disturbances of voluntary movements as seen in Parkinson’s disease and dystonia. It is essential to understand neural activity changes in the cortico-basal ganglia networks in these movement disorders, because it can explain pathophysiological mechanisms for their symptoms in human patients. In Parkinson’s disease, signals through the hyperdirect and indirect pathways are enhanced, and signals through the direct pathway are reduced in both spatial and temporal domains. Thus, intended motor programs cannot be released at an appropriate timing, causing akinesia (difficulty to initiate movements). On the other hand, in dystonia, the direct pathway dominates over the hyperdirect and indirect pathways spatiotemporally. Tiny excitation in the cerebral cortex releases unintended movements randomly, causing involuntary movements. Therefore, these movement disorders can be considered as network disorders.
Stereotactic surgery, making a small lesion or applying high-frequency electrical stimulation (deep brain stimulation, DBS), in the basal ganglia ameliorates symptoms of movement disorders. GPi-DBS inhibits cortically induced responses and spontaneous discharges in the GPi by strong GABAergic inhibition, suggesting that GPi-DBS blocks information flow through the GPi. Both lesion and DBS block abnormal information flow from the basal ganglia to the thalamus and cortex, and suppress the expression of abnormal motor symptoms. The beneficial mechanism of stereotactic surgery can also be explained by the notion of network disorders.

1. Nambu, A., 2008. Seven problems on the basal ganglia. Current Opinion in Neurobiology, 18(6), pp.595–604.
2. Nambu, A., Tachibana, Y. & Chiken, S., 2015. Cause of parkinsonian symptoms: Firing rate, firing pattern or dynamic activity changes? Basal Ganglia, 5(1), pp.1–6.
3. Chiken, S. & Nambu, A., 2016. Mechanism of Deep Brain Stimulation. The Neuroscientist, 22(3), pp.313–322.
4. Ozaki, M. et al., 2017. Optogenetic Activation of the Sensorimotor Cortex Reveals “Local Inhibitory and Global Excitatory” Inputs to the Basal Ganglia. Cerebral Cortex, 27(12), pp.5716–5726.
5. Iwamuro, H. et al., 2017. Information processing from the motor cortices to the subthalamic nucleus and globus pallidus and their somatotopic organizations revealed electrophysiologically in monkeys. European Journal of Neuroscience, 46(11), pp.2684–2701.
6. Nambu, A. et al., 2011. Reduced Pallidal Output Causes Dystonia. Frontiers in Systems Neuroscience, 5(November), pp.1–6.

Sang Wan Lee, Ph.D.

Reinforcement learning (RL) has demonstrated an ability to succeed in a few arduous tasks, emerging as a general framework for decision making in neuroscience and robotics. This talk introduces our research team’s twofold approach to better understanding the nature of human RL. The first part of the talk focuses on the prefrontal-striatal circuitry for meta RL. By using a combination of model-based experimental design and computational modelling, I will discuss the structure of prefrontal-striatal network for meta RL. I will then discuss the key variables that guide this process: prediction error, uncertainty, self-expectation, task complexity, and metacognitive ability. These evidences accumulate to suggest a theoretical idea about how the meta RL resolves tradeoff issues: performance-efficiency, speed-accuracy, and exploration-exploitation. The last part of the talk outlines a more pragmatic approach to improving optimality of human RL. A detailed insight into these issues not only permits advances in a reinforcement learning theory, but also helps us understand the nature of human intelligence on a deeper level.

1. Lee, S.W., Shimojo, S. & O’Doherty, J.P., 2014. Neural computations underlying arbitration between model-based and model-free learning. Neuron, 81(3), pp.687–699.
2. Lee, S.W., O’Doherty, J.P. & Shimojo, S., 2015. Neural computations mediating one-shot learning in the human brain. PLOS Biology, 13(4), p.e1002137.
3. Lee, J.H. et al., 2019. Towards high performance, memory efficient, and fast reinforcement learning - lessons from decision neuroscience. Science Robotics, 4(26).

Tetsuo Yamamori, Ph.D.

I started my scientific career by studying bacterial molecular genetics when I was a graduate student at the department of biophysics in Kyoto University under supervise of Dr. Takashi Yura. There, I eventually found a phenomenon called Heat Shock in bacteria (Escherichia coli). Heat Shock was originally found as a rapid change of chromatin structure (puff) upon an exposure to high temperature (Ashburner, 1972). This phenomenon was later found to be linked to a new set of RNA and protein synthesis (Ashburner & Bonner, 1979). When I was analyzing temperature sensitive mutants in E. coli, I observed that a set of several genes were rapidly induced when the culture temperature was elevated from 30 to 42 oC (Yamamori et al.1978). I thought that this phenomenon may be analogous to the Drosophila Heat Shock. Indeed, it turned out that the proteins induced by heat shock are conserved from bacteria to human. I also found a gene that regulates the induction of the heat shock proteins at a transcriptional level (Yamamori & Yura, 1982). These two findings established Heat Shock as a universal (very conserved) phenomena that are genetically controlled throughout living organisms from bacteria to Humans.
The discovery of the bacterial Heat Shock led me to be interested in the nervous system, in which a group of organisms, i.e., animals, have systematically evolved to the outside environment by having their energy source from the other organisms or eating. At that time when I received Ph D (1981), it was not easy to find a laboratory in which a postdoc who was trained in molecular biology was able to study the nervous system. It was about the time that the concept of multidisciplinary Neuroscience was about to emerge. I therefore took an opportunity to study a cell line system for a neural and glial differentiation at the University of Colorado, Boulder, under the supervise of Dr. Noboru Sueoka in 1981. During my study in Boulder, I was kindly asked by Dr. Naka-akira Tsukahara, a well-known physiologist by sprouting in the red nucleus (Tsukahara, 1981), if I was interested in working in his laboratory in Osaka University. However, to my regret, Dr. Tsukahara was killed by the JAL 123 accident in 1985. I then started to work with Dr. Paul Patterson at Cal Tech for a protein factor that switch adrenergic sympathetic neurons to cholinergic neurons. I found this cholinergic differentiation factor (CDF) as Leukemia Inhibitory Factor (Yamamori et al., 1989).
Just after the CDF paper was published, Dr. Masao Ito started a new research system called “Frontier Research System” in RIKEN, Wako. Dr. Ito has been known by his finding of long term depression (LTD) in the cerebellum. LTD had been proposed as the key mechanism for cerebellar learning by Albus (1971) and Marr (1969), but never had been proved until Dr. Ito’s team first experimentally found it (Ito, Sakurai & Tongroach, 1982). I was asked if I am interested in working in his team. I thought that it would be a good opportunity to learn system neuroscience and joined his team in 1991. I studied the expression of immediate early genes and protein synthesis that are possibly linked with cerebellar LTD (Nakazawa et al.1993; Karachot et al. 2000, 2001).
In 1994, I was appointed as a professor at National Institute of Basic Biology, Okazaki, high throughput sequence was not possible at that time, and therefore we first used differential display method, and then the RLCS (restriction land mark cDNA scanning) method to identify the species of mRNA that were selectively expressed representative cortical areas of macaque monkeys. We have published a series of papers and finally concluded that there are two groups of genes that are highly expressed in cortical areas of adult macaques: One group is high in primary sensory areas particularly in the primary visual cortex and the other is high in association areas and frontal cortex (Yamamori, 2011). Although we found these genes selectively expressed in the primate neocortex, it was difficult to know the function of the genes in the macaque cortex except for one case of serotonin receptor subtype in which specific agonists and antagonists of serotonin receptors were available (Watakabe et al. 2009). There was no general method that we could manipulate genes in the primate cortex. I therefore wanted to explore a system that we can identify the functions of the genes in the primate cortex. Dr. Kathleen Rockland, a team leader in RKIEN at that time and a system neuroanatomist, suggested me to pay attention on the marmoset, New World monkey. I started to learn marmosets for the possible application for the study of gene functions in the primate cortex. Supported by a grant of Strategic Research Program for Brain Sciences (Project C: from 2008 to 2012) led by Dr. Tadashi Isa, National Institute for Physiological Sciences and currently Kyoto University, I started to establish a system to examine gene functions in the marmoset cortex. In the core members of the same Project C, Drs. Erica Sasaki’s (Central Insitute for Experimental Animals) and Hideyuki Okano’s (Keio Medical School) groups reported the first germline transmittable transgenic line (2009). I myself focused on using virus vectors to visualize and manipulate genes in the marmoset cortex. We were able to observe neural activities by Ca imaging in the marmoset cortex (Sadakene et al. 2015). We were also able knock down the mRNA of Dopamine receptor 1 (D1R) and D2R and observed the behavioral effects on the marmoset caudate (Takaji et al., 2016).
Using virus vector mediated fluorescent proteins, we are currently mapping a large-scale marmoset cortical projection in collaborations including Dr. Shin Ishi and his colleagues as Brain/MINDs project in Japan. In my talk, I would present some of our recent data as well as the main scientific results as I describe above throughout my scientific career.

1. Okano, H. et al., 2016. Brain/MINDS: A Japanese National Brain Project for Marmoset Neuroscience. Neuron, 92(3), pp.582–590.
2. Yamamori, T., 2011. Selective gene expression in regions of primate neocortex: Implications for cortical specialization. Progress in Neurobiology, 94(3), pp.201–222.
3. Sadakane, O. et al., 2015. Long-Term Two-Photon Calcium Imaging of Neuronal Populations with Subcellular Resolution in Adult Non-human Primates. Cell Reports, 13(9), pp.1989–1999.
4. Yamamori, T. et al., 1989. The cholinergic neuronal differentiation factor from heart cells is identical to leukemia inhibitory factor. Science, 246(4936), pp.1412–1416.
5. Yamamori, T. et al., 1978. Transient regulation of protein synthesis in Escherichia coli upon shift-up of growth temperature. Journal of Bacteriology, 134(3), pp.1133–1140.
6. Yamamori, T. & Yura, T., 1982. Genetic control of heat-shock protein synthesis and its bearing on growth and thermal resistance in Escherichia coli K-12. Proceedings of the National Academy of Sciences, 79(3), pp.860–864.