The Allen Institute for Brain Science is a Seattle-based independent, nonprofit medical research organization dedicated to accelerating the understanding of how the human brain works. The Allen Institute promotes the advance of brain research by providing free data and tools to scientists worldwide with the aim of catalyzing discovery in disparate research programs and disease areas.Started with $100 million in seed money from philanthropist Paul Allen in 2003, the Institute tackles projects at the leading edge of science—far-reaching projects at the intersection of biology and technology. The resulting data create free, publicly available resources that fuel discovery for countless researchers. Wikipedia.
News Article | May 11, 2017
The newly announced scientific board includes David Eagleman, PhD, Professor of Neuroscience, Stanford University, Founder & CSO, BrainCheck, and cofounder & CSO, NeoSensory; Christof Koch, PhD, CSO & President of Allen Institute for Brain Science; Emmanuel Mignot, MD, PhD, Director of Stanford Sleep Center; and Cedric Villani, PhD, winner of the 2010 Fields Medal and Director of Henri Poincarré Institute. These experts will help Rythm define its research strategy, encompassing sleep diagnosis, treatment, improvement, and understanding of the human brain. The human brain is the most complex organ in the known universe, and understanding the brain and sleep are among the biggest scientific challenges of the century. One-third of the population does not sleep well or sufficiently long. Currently, there are no obvious solutions to this epidemic. However, international research in neuroscience and Rythm are at the forefront of developing non-invasive solutions that are accurate and effective. The company is tackling multiple, ambitious challenges at the nexus of multiple disciplines: medicine, neurosciences, and mathematics. In forming a board, Rythm successfully sought to attract the best-of-the-best from across disciplines who are aligned with the company's mission and vision. "The brain is the most complex system known today, and within that field of study, sleep is a new domain that presents a variety of complex problems and solutions," said Hugo Mercier, CEO & Co-founder at Rythm. "The diversity of fields, experience, and intellect that the members of our board bring will help Rythm unlock major challenges and pursue the right research directions." David Eagleman is a professor at Stanford University in the department of Psychiatry & Behavioral Sciences, known for his work on brain plasticity, time perception, synesthesia, and neurolaw. He also serves as the Director of the Center for Science and Law. He is a Guggenheim Fellow, a council member in the World Economic Forum, and a popular TED speaker. He has launched two companies from his laboratory: NeoSensory and BrainCheck. He is a New York Times bestselling author published in 31 languages and is the writer and creator of the Emmy-nominated TV series, The Brain with David Eagleman. He is a renowned scientist with articles in all the major academic journals and profiles in national magazines such as the New Yorker. He is a regular commenter on national television and radio. "We don't yet fully understand why we sleep and dream, but we're aware that it's related to the consolidation of learning and memory," said Eagleman. "I am excited to work with Rythm to better unmask the mysteries and nuances of sleep state, and to be able to leverage that understanding to improve lives. Inadequate sleep prevents people from reaching their full potential. The improvement of sleep opens the hope of functioning at a more optimized mental performance." Christof Koch is a physicist turned neuroscientist serving as the President and Chief Scientific Officer of the Allen Institute for Brain Science in Seattle. He is leading a team of 300 scientists, engineers and staff engaged in a ten-year project that aims to understand the building blocks of the mammalian brain. Koch previously served as a professor at the California Institute of Technology for nearly 30 years, specializing in the biophysics of the brain and the neural bases of consciousness, and has been influential in arguing that consciousness can be approached using the modern tools of neurobiology. As a member of Rythm's board, Koch will contribute his neuroscientific expertise on how sleep relates to the brain and its electrical behavior in health and disease. "We have so much more to learn about the relationship that sleep has to the functioning of our brains and our health," said Koch. "Working with Rythm is an opportunity to bring academia together with the development of consumer products so that sleep research can become practicable." Emmanuel Mignot is the Craig Reynolds Professor of Sleep Medicine at Stanford Medical School and the Director of the Stanford Center for Sleep Sciences and Medicine. He is a recognized authority on sleep research and medicine, known primarily for his work on narcolepsy. He is a member of the National Academies of Sciences and Medicine and has received numerous research grants and honors, including National Institute of Health Research, Howard Hughes Medical Institute Investigator and McKnight Neuroscience awards. He is the co-author of more than 200 original scientific publications, and serves on the editorial board of scientific journals in the field of sleep and biology research. He formerly served as the president of the Sleep Research society, chair of the National Center on Sleep Disorders Research advisory board of the National institutes of Health, and chair of the Board of Scientific Counselors of the National Institute of Mental Health. "I've always been intrigued by the enigma of sleep and devoted my career to studying sleep disorders," said Mignot. "With the rapid growth of portable technology, biology and analytics, it is an exciting time for sleep, with plenty of opportunities to increase well-being. Rythm is bringing together people spanning multiple fields of science and technology to push forward our understanding of sleep and improve the diagnosis of sleep disorders. I look forward to contributing my knowledge and working with Rythm to help realize this vision." Cedric Villani is one of the world's most famous mathematicians who was awarded the Fields Medal, the world's most prestigious math award, in 2010. Currently, he is a professor at the University of Lyon and serves as the director of Pierre and Marie Curie University 's Institut Henri Poincaré. Villani's work focuses on partial differential equations, Riemannian geometry and mathematical physics. He received the Fields Medal for his work on Landau damping and the Boltzmann equation. He is also a well-known author and speaker, renowned for his passionate ability to make math exciting and accessible. Villani's expertise on computational mathematics and machine learning will be a valuable asset to Rythm because both areas are critical to the understanding of sleep and the brain. Until recently, machine learning mimicked brain functions, but now the new frontier is to understand how the brain works, and sleep represents an ideal entry door. This world class team serves as a validation of all the important work Rythm has done since 2014. The company is leveraging advancements in neuroscience, neurotechnology, artificial intelligence, and mathematics to propel sleep research and medicine forward and bring a real sleep solution to the market. This unique solution will introduce a whole new category of product that is efficient but non-invasive, and this demands a strong research effort and the development of sophisticated technology. The board is not only working with Rythm on diagnosis and treatment, but also to help build the product that will launch in Summer 2017. Rythm is a leading neurotechnology company. Bringing together the world's foremost experts in hardware, software, and neuroscience, Rythm builds consumer technology that stimulates brain function to enhance individual health and performance. The company's first product, Dreem, is sleep solution that uses brain activity and sound stimulation to increase the quality of sleep in a non-invasive way. Based in Paris and San Francisco, Rythm has raised substantial funding from investors, awards and government grants to support a world-class team of more than 70 people. For more information, visit www.dreem.com. To view the original version on PR Newswire, visit:http://www.prnewswire.com/news-releases/rythm-accelerates-sleep-research-and-neurotechnology-efforts-with-a-prestigious-scientific-advisory-board-and-advances-in-the-ai-xprize-competition-300456427.html
News Article | May 17, 2017
Most research institutions are essentially collections of independent laboratories, each run by principal investigators who head a team of trainees. This scheme has ancient roots and a track record of success. But it is not the only way to do science. Indeed, for much of modern biomedical research, the traditional organization has become limiting. A different model is thriving at the Broad Institute of MIT and Harvard in Cambridge, Massachusetts, where I work. In the 1990s, the Whitehead Institute for Biomedical Research, a self-governing organization in Cambridge affiliated with the Massachusetts Institute of Technology (MIT), became the academic leader in the Human Genome Project. This meant inventing and applying methods to generate highly accurate DNA sequences, characterize errors precisely and analyse the outpouring of data. These project types do not fit neatly into individual doctoral theses. Hence, the institute created a central role for staff scientists — individuals charged with accomplishing large, creative and ambitious projects, including inventing the means to do so. These non-faculty scientists work alongside faculty members and their teams in collaborative groups. When leaders from the Whitehead helped to launch the Broad Institute in 2004, they continued this model. Today, our work at the Broad would be unthinkable without professional staff scientists — biologists, chemists, data scientists, statisticians and engineers. These researchers are not pursuing a tenured academic post and do not supervise graduate students, but do cooperate on and lead projects that could not be accomplished by a single academic laboratory. Physics long ago saw the need to expand into different organizational models. The Manhattan Project, which during the Second World War harnessed nuclear energy for the atomic bomb, was not powered by graduate students. Europe's particle-physics laboratory, CERN, does not operate as atomized labs with each investigator pursuing his or her own questions. And the Jet Propulsion Laboratory at the California Institute of Technology in Pasadena relies on professional scientists to get spacecraft to Mars. In biology, many institutes in addition to the Broad are experimenting with new organizational principles. The Mechanobiology Institute in Singapore pushes its scientists to use tools from other disciplines by discouraging individual laboratories from owning expensive equipment unless it is shared by all. The Howard Hughes Medical Institute's Janelia Research Campus in Ashburn, Virginia, the Salk Institute of Biological Sciences in La Jolla, California, and the Allen Institute for Brain Science in Seattle, Washington, effectively mix the work of faculty members and staff scientists. Disease-advocacy organizations, such as the ALS Therapy Development Institute in Cambridge, do their own research without any faculty members at all. Each of these institutes has a unique mandate, and many are fortunate in having deep resources. They also had to be willing to break with tradition and overcome cultural barriers. At famed research facilities of yore, such as Bell Labs and IBM Laboratories, the title 'staff scientist' was a badge of honour. Yet to some biologists the term suggests a permanent postdoc or senior technician — someone with no opportunities for advancement who works solely in a supervisor's laboratory, or who runs a core facility providing straightforward services. That characterization sells short the potential of professional scientists. The approximately 430 staff scientists at the Broad Institute develop cutting-edge computational methods, invent and incorporate new processes into research pipelines and pilot and optimize methodologies. They also transform initial hits from drug screens into promising chemical compounds and advance techniques to analyse huge data sets. In summary, they chart the path to answering complex scientific questions. Although the work of staff scientists at the Broad Institute is sometimes covered by charging fees to its other labs, our faculty members would never just drop samples off with a billing code and wait for data to be delivered. Instead, they sit down with staff scientists to discuss whether there is an interesting collaboration to be had and to seek advice on project design. Indeed, staff scientists often initiate collaborations. Naturally, tensions still arise. They can play out in many ways, from concerns over how fees are structured, to questions about authorship. Resolving these requires effort, and it is a task that will never definitively be finished. In my view, however, the staff-scientist model is a win for all involved. Complex scientific projects advance more surely and swiftly, and faculty members can address questions that would otherwise be out of reach. This model empowers non-faculty scientists to make independent, creative contributions, such as pioneering new algorithms or advancing technologies. There is still much to do, however. We are working to ensure that staff scientists can continue to advance their careers, mentor others and help to guide the scientific direction of the institute. As the traditional barriers break down, science benefits. Technologies that originate in a faculty member's lab sometimes attract more collaborations than one laboratory could sustain. Platforms run by staff scientists can incorporate, disseminate and advance these technologies to capture more of their potential. For example, the Broad Institute's Genetic Perturbation Platform, run by physical chemist David Root, has honed high-throughput methods for RNA interference and CRISPR screens so that they can be used across the genome in diverse biological contexts. Staff scientists make the faculty more productive through expert support, creativity, added capacity and even mentoring in such matters as the best use of new technologies. The reverse is also true: faculty members help staff scientists to gain impact. Our staff scientists regularly win scientific prizes and are invited to give keynote lectures. They apply for grants as both collaborators and independent investigators, and publish regularly. Since 2011, staff scientists have led 36% of all the federal grants awarded for research projects at the Broad Institute (see ‘Staff-led grants’). One of our staff scientists, genomicist Stacey Gabriel, topped Thomson Reuters' citation analysis of the World's Most Influential Scientific Minds in 2016. She co-authored 25 of the most highly cited papers in 2015 — a fact that illustrates both how collaborative the Broad is and how central genome-analysis technologies are to answering key biological questions. At the Broad Institute's Stanley Center for Psychiatric Research, which I direct, staff scientists built and operate HAIL, a powerful open-source tool for analysis of massive genetics data sets. By decreasing computational time, HAIL has made many tasks 10 times faster, and some 100 times faster. Staff scientist Joshua Levin has developed and perfected RNA-sequencing methods used by many colleagues to analyse models of autism spectrum disorders and much else. Nick Patterson, a mathematician and computational biologist at the Stanley Center, began his career by cracking codes for the British government during the cold war. Today, he uses DNA to trace past migrations of entire civilizations, helps to solve difficult computational problems and is a highly valued support for many biologists. Why haven't more research institutions expanded the roles of staff scientists? One reason is that they can be hard to pay for, especially by conventional means. Some funding agencies look askance at supporting this class of professionals; after all, graduate students and postdocs are paid much less. In my years leading the US National Institute of Mental Health, I encountered people in funding bodies across the world who saw a rising ratio of staff to faculty members or of staff to students as evidence of fat in the system. That said, there are signs of flexibility. In 2015, the US National Cancer Institute began awarding 'research specialist' grants — a limited, tentative effort designed in part to provide opportunities for staff scientists. Sceptical funders should remember that trainees often take years to become productive. More importantly, institutions' misuse of graduates and postdocs as cheap labour is coming under increasing criticism (see, for example, et al. Proc. Natl Acad. Sci. USA 111, 5773–5777; 2014). Faculty resistance is also a factor. I served as Harvard University's provost (or chief academic officer) for a decade. Several years in, I launched discussions aimed at expanding roles for staff scientists. Several faculty members worried openly about competition for space and other scarce resources, especially if staff scientists were awarded grants but had no teaching responsibilities. Many recoiled from any trappings of corporatism or from changes that felt like an encroachment on their decision-making. Some were explicitly concerned about a loss of access and control, and were not aware of the degree to which staff scientists' technological expertise and cross-disciplinary training could help to answer their research questions. Institutional leaders can mitigate these concerns by ensuring that staff positions match the shared goals of the faculty — for scientific output, education and training. They must explain how staff-scientist positions create synergies rather than silos. Above all, hiring plans must be developed collaboratively with faculty members, not by administrators alone. The Broad Institute attracts world-class scientists, as both faculty members and staff. Its appeal has much to do with how staff scientists enable access to advanced technology, and a collaborative culture that makes possible large-scale projects rarely found in academia. The Broad is unusual — all faculty members also have appointments at Harvard University, MIT or Harvard-affiliated hospitals. The institute has also benefited from generous philanthropy from individuals and foundations that share our values and believe in our scientific mission. Although traditional academic labs have been and continue to be very productive, research institutions should look critically and creatively at their staffing. Creating a structure like that of the Broad Institute would be challenging in a conventional university. Still, I believe any institution that is near an academic health centre or that has significant needs for advanced technology could benefit from and sustain the careers of staff scientists. If adopted judiciously, these positions would enable institutions to take on projects of unprecedented scope and scale. It would also create a much-needed set of highly rewarding jobs for the rising crop of talented researchers, particularly people who love science and technology but who do not want to pursue increasingly scarce faculty positions. A scientific organization should be moulded to the needs of science, rather than constrained by organizational traditions.
News Article | May 24, 2017
GAD2-IRES-Cre, VGLUT2-IRES-Cre, VGAT-IRES-Cre, GAD1-eGFP, CCK-IRES-Cre, CRH-IRES-Cre, TAC1-IRES-Cre, and HDC-IRES-Cre mice (Jackson stock numbers 010802, 016963, 016962, 007677, 012706, 012704, 021877, and 021198, respectively) were obtained from Jackson Laboratory19, 32, 33 and VGLUT2-eGFP mice were from MMRRC (MMRRC 011835-UCD). PDYN-IRES-Cre mice were obtained from Bradford Lowell34. GAL-Cre mice were obtained from GENSAT (stock number KI87). Mice were housed in 12 h light–dark cycle (lights on 7:00 and off at 19:00) with free access to food and water. Experiments were performed in adult male or female mice (6–12 weeks old). All procedures were approved by Institutional Animal Care and Use Committees of the University of California, Berkeley, University of California, San Francisco, Allen Institute for Brain Science, and Stanford University and were done in accordance with federal regulations and guidelines on animal experimentation. Note that, in different experiments, GAD1, GAD2, and VGAT were used to identify GABAergic neurons. To examine the relationship between GAD1, GAD2, and VGAT in the POA, we quantified the overlap between GAD1 and VGAT on the basis of the double in situ hybridization data from the Allen Mouse Brain Atlas (http://connectivity.brain-map.org/transgenic/experiment/100142488) and found that 95% (327 out of 345) of VGAT-positive neurons also contained GAD1. Comparison between GAD1 and GAD2 expression in the POA (http://connectivity.brain-map.org/transgenic/experiment/100142491) showed that 99% (246 out of 248) of GAD1 neurons also contained GAD2. Together, these data indicate a very high degree of overlap between GAD1, GAD2, and VGAT in the POA. AAV -EF1α-DIO-ChR2–eYFP and AAV -hSyn-FLEX-hM4D(Gi)–mCherry were obtained from the University of North Carolina vector core. The final titre was estimated to be ~1012 genome copies per millilitre. AAV -EF1α-DIO-ChR2–eYFP, AAV -EF1α-DIO–eYFP, AAV -EF1α-DIO-iC++–eYFP, and AAV -EF1α-DIO-iC++–eYFP were obtained from Stanford University virus core. Lentivirus rEIAV-DIO-TLoop-ChR2–eYFP, rEIAV-DIO-TLoop-iC++–eYFP, and rEIAV-DIO-TLoop-nls–eYFP were obtained from Salk virus core and Allen Institute for Brain Science13. Rabies-tracing reagents (AAV-CAG-FLExloxP-TVA–mCherry, AAV-CAG-FLExloxP-RG (RG, rabies glycoprotein) and EnvA-pseudotyped, rabies-glycoprotein-deleted, and GFP-expressing rabies viral particles (RVdG)), cTRIO reagents (CAV-FLExloxP-Flp, AAV-FLExFRT-TVA–mCherry, AAV-FLExFRT-RG, and EnvA-pseudotyped, rabies-glycoprotein-deleted, and GFP-expressing rabies viral particles (RVdG)) and axon arborization analysis reagents (CAV-FLExloxP-Flp and AAV-hSyn1-FLExFRT–mGFP–2A-synaptophysin–mRuby) were obtained from Stanford University15. HSV-LoxSTOPLox-FlagHA-L10a was obtained from the University of California, San Francisco. Mice of a specific genotype were randomly assigned to experimental and control groups. Experimental and control animals were subjected to exactly the same surgical and behavioural manipulations. Data from animals used in experiments were excluded on the basis of histological criteria that included injection sites, virus expression, and optical fibre placement. Only animals with injection sites and optic fibre placement in the region of interest were included. To implant EEG and EMG recording electrodes, adult mice (6–12 weeks old) were anaesthetized with 1.5–2% isoflurane and placed on a stereotaxic frame. Two stainless steel screws were inserted into the skull 1.5 mm from midline and 1.5 mm anterior to the bregma, and two others were inserted 3 mm from midline and 3.5 mm posterior to the bregma. Two EMG electrodes were inserted into the neck musculature. Insulated leads from the EEG and EMG electrodes were soldered to a 2 × 3-pin header, which was secured to the skull using dental cement. For optogenetic activation/inhibition experiments, a craniotomy was made on top of the target region for optogenetic manipulation in the same surgery as for EEG and EMG implant, and 0.1–0.5 μl virus was injected into the target region using Nanoject II (Drummond Scientific) via a micropipette. We then implanted optic fibres bilaterally into the target region. Dental cement was applied to cover the exposed skull completely and to secure the implants for EEG and EMG recordings to the screws. After surgery, mice were allowed to recover for at least 2–3 weeks before experiments. For anti-histamine experiments (Extended Data Fig. 4), triprolidine (Tocris) was administered intraperitoneally at 20 mg per kg (body weight) and brain states were recorded for 3 h. For retrograde tracing in Extended Data Fig. 1a, 0.2–0.3 μl red or green RetroBeads (Lumafluor) was injected into each target region. For optrode recording experiments, the optrode assembly was inserted into the POA at a depth of 4.9 mm. Screws were attached to the skull for EEG recordings, and an EMG electrode was inserted into the neck musculature. The optrode assembly, screws, and EEG/EMG electrodes were secured to the skull using dental cement. These procedures are related to the results in Fig. 3 and Extended Data Fig. 6. For rabies tracing, AAV-CAG-FLExloxP-TVA–mCherry and AAV-CAG-FLExloxP-RG were injected into the TMN of HDC-Cre mice. Two to three weeks later, EnvA-pseudotyped, glycoprotein-deleted, and GFP-expressing rabies viral particles (RVdG) were injected into the TMN, and mice were euthanized 1 week later. These procedures are related to the results in Extended Data Fig. 2. For cTRIO experiments, a retrograde virus CAV-FLExloxP-Flp (5.0 × 1012 genome copies per millilitre) was injected into either the TMN or the PFC of GAD2-Cre mice to express Flp recombinase specifically in GABAPOA→TMN or GABAPOA→PFC neurons, and AAV-FLExFRT-TVA–mCherry (2.6 × 1012 genome copies per millilitre) and AAV-FLExFRT-RG (1.3 × 1012 genome copies per millilitre) were injected into the POA to express TVA (the receptor for the EnvA envelope glycoprotein)–mCherry and rabies glycoprotein in the Flp-expressing neurons. Two to three weeks later, EnvA-pseudotyped, glycoprotein-deleted, and GFP-expressing rabies viral particles (RVdG) (5.0 × 108 colony forming units per millilitre) were injected into the POA, and mice were euthanized 1 week later for histology. These procedures are related to the results in Extended Data Fig. 7. For axon arborization experiments, CAV-FLExloxP-Flp was injected into TMN, and AAV-hSyn1-FLExFRT–mGFP–2A-synaptophysin–mRuby was injected into the POA of GAD2-Cre mice. Mice were euthanized 4–7 weeks later for histology. These procedures are related to the results in Extended Data Fig. 7. For pharmacogenetic experiments, AAV -hSyn-FLEX-hM4D(Gi)–mCherry was injected bilaterally into the POA. These procedures are related to the results in Extended Data Fig. 10. For TRAP experiments, we injected Cre-inducible HSV expressing the large ribosomal subunit protein Rpl10a fused with Flag/haemagglutinin tag (HSV-LoxSTOPLox-FlagHA-L10a) into the TMN of VGAT-Cre mice. After 30–45 days of expression, the POA was dissected, and ribosome immunoprecipitation was performed to pull down the messenger RNAs (mRNAs) attached to Rpl10a. These procedures are related to the results in Fig. 4 and Extended Data Fig. 8. For single-cell RNA-seq experiments, rEIAV-DIO-TLoop-nls–eYFP was injected into the TMN of GAD2-Cre and VGAT-Cre mice. Four weeks later, we dissociated eYFP-labelled POA neurons for single-cell RNA-seq. These procedures are related to the results in Fig. 4 and Extended Data Fig. 8. For immunohistochemistry-detecting peptides, mice received a single intraventricular injection of colchicine (12 μg) 1–2 days before killing. These procedures are related to the results in Fig. 4. The stereotaxic coordinates were as follows. TMN: anteroposterior (AP) −2.45 mm, mediolateral (ML) 1 mm, dorsoventral (DV) 5–5.2 mm from the cortical surface; POA: AP 0 mm, ML 0.7 mm, DV 5.2 mm; PFC: AP +2.0 mm, ML 0.4 mm, DV 2 mm; vlPAG: AP −4.7 mm, ML 0.7 mm, DV 2.3 mm; dorsomedial hypothalamus: AP −1.8 mm, ML 0.4 mm, DV 5.2 mm; habenula: AP −1.8 mm, ML 0.5 mm, DV 2.2 mm. Sleep deprivation started at the beginning of the light period (7:00) and lasted till 13:00. Mice were kept awake by a combination of cage tapping, introduction of foreign objects such as paper towels, cage rotation, and fur stroking with a paintbrush35, gentle handling procedures that have been used extensively to induce sleep deprivation36. EEG and EMG were not recorded during sleep deprivation and recovery. After 6 h of deprivation, sleep-deprived mice were allowed rebound sleep for 4 h before being euthanized by cervical dislocation and decapitation. c-Fos immunohistochemistry was performed as described below. These procedures are related to the results in Extended Data Figs 1 and 2. Behavioural experiments were performed in home cages placed in sound-attenuating boxes. Sleep recordings were performed between 12:00 and 19:00 (light on at 7:00 and off at 19:00). EEG and EMG electrodes were connected to flexible recording cables via a mini-connector. EEG and EMG signals were recorded and amplified using AM Systems, digitally filtered (0.1–1,000 Hz and 10–1,000 Hz for EEG and EMG recordings respectively), and digitized at 600 Hz using LabView. Spectral analysis was performed using fast Fourier transform, and brain states were classified into NREM, REM, and wake states (wake: desynchronized EEG and high EMG activity; NREM: synchronized EEG with high-amplitude, low-frequency (0.5–4 Hz) activity and low EMG activity; REM: high power at theta frequencies (6–9 Hz) and low EMG activity). Brain states were classified into NREM sleep, REM sleep, and wakefulness using custom-written MATLAB software, and the classification was performed without any information about the identity of the animal or laser stimulation timing as previously described25. Each optic fibre (200 μm diameter; ThorLabs) was attached through an FC/PC adaptor to a 473-nm blue laser diode (Shanghai laser), and light pulses were generated using a Master 8 (A.M.P.I.). All photostimulation/inhibition experiments were conducted bilaterally and fibre optic cables were connected at least 2 h before the experiments for habituation. For photostimulation/inhibition experiments in ChR2-, iC++-, or eYFP-expressing mice, light pulses (10 ms per pulse, 10 Hz, 4–8 mW) or step pulses (60 s) were triggered using Master 8 that provided simultaneous input into two blue lasers. In each optogenetic manipulation experiment, inter-stimulation interval for optogenetic manipulation was chosen randomly from a uniform distribution between 15 and 25 min. Custom-made optrodes37 consisted of an optic fibre (200 μm in diameter) glued together with six pairs of stereotrodes. Two FeNiCr wires (Stablohm 675, California Fine Wire) were twisted together and electroplated to an impedance of ~ 600 kΩ using a custom-built plating device. The optrode was attached to a driver to allow vertical movement of the optrode assembly. The optrode was slowly lowered to search for light-responsive neurons. Wires to record cortical EEG and EMG from neck musculatures were also attached for simultaneous recordings. A TDT RZ5 amplifier was used for all the recordings, signals were filtered (0.3–8 kHz) and digitized at 25 kHz. At the end of the experiment, an electrolytic lesion was made by passing a current (100 μA, 10 s) through one or two electrodes to identify the end of the recording tract. Spikes were sorted offline on the basis of the waveform energy and the first three principal components of the spike waveform on each stereotrode channel. For single unit isolation, all channels were separated into groups and spike waveforms were identified either manually using Klusters (http://neurosuite.sourceforge.net/) or automatically using the software klustakwik (http://klustakwik.sourceforge.net/). The quality of each unit was assessed by the presence of a refractory period and quantified using isolation distance and L . Units with an isolation distance <20 and L >0.1 were discarded38. To identify ChR2-tagged neurons, laser pulse trains (10 and/or 20 Hz) were delivered intermittently every minute. A unit was identified as ChR2-expressing if spikes were evoked by laser pulses with short first-spike latency (<6 ms for all units in our sample) and the waveforms of the laser-evoked and spontaneous spikes were highly similar (correlation coefficient >0.9). Mean latency of all identified units was 3.05 ms. Mean correlation coefficient of all identified units was 0.99. To calculate the average firing rate of each unit in each brain state, spikes during the laser pulse trains were excluded. These procedures are related to the results in Fig. 3 and Extended Data Fig. 6. Mice were deeply anaesthetized and transcardially perfused using PBS buffer followed by 4% paraformaldehyde in PBS. Brains were post-fixed in fixative and stored in 30% sucrose in PBS overnight for cryoprotection. Brains were embedded and mounted with Tissue-Tek OCT compound (Sakura Finetek) and 20 μm sections were cut using a cryostat (Leica). Brain slices were washed using PBS, permeabilized using PBST (0.3% Triton X-100 in PBS) for 30 min and then incubated with blocking solution (5% normal goat serum or normal donkey serum in PBST) for 1 h followed by primary antibody incubation overnight at 4 °C using the following antibodies: anti–GFP antibody (A-11122 or A-11120, Life technologies, 1:1,000); anti-cFos antibody (sc-52-G and sc-52, Santa Cruz Biotech, 1:1,000); anti-CCK-8 antibody (20078, Immunostar, 1:500); anti-CRH antibody (sc-1759, Santa Cruz Biotech, 1:500); anti-haemagglutinin antibody (C29F4, Cell Signaling tech, 1:1,000); and anti-HDC antibody (16045, Progen, 1:1,000). The next day, slices were washed with PBS and incubated with appropriate secondary antibodies for 2 h (1:500, all from Invitrogen): A-11008, Alexa Fluor 488 goat anti-rabbit IgG; A-21206, Alexa Fluor 488 donkey anti-rabbit IgG; A-11055, Alexa Fluor 488 donkey anti-goat IgG; A-21202, Alexa Fluor 488 donkey anti-mouse IgG; A-11012, Alexa Fluor 594 goat anti-rabbit IgG; A-21207, Alexa Fluor 594 donkey anti-rabbit IgG; A-11058, Alexa Fluor 594 donkey anti-goat IgG; A-21245, Alexa Fluor 647 goat anti-rabbit IgG. The slices were washed with PBS followed by counterstaining with DAPI or Hoechst and coverslipped. Fluorescence images were taken using a confocal microscope (LSM 710 AxioObserver Inverted 34-Channel Confocal, Zeiss) or Nanozoomer (Hamamatsu). FISH was performed with two methods. First, FISH for CCK, CRH, TAC1, and GAD1 was done using RNAscope assays according to the manufacturer’s instructions (Advanced Cell Diagnostics). Second, to make TAC1, GAD1, and GAD2 FISH probes, DNA fragments containing the coding or untranslated sequences were amplified using PCR from mouse whole brain complementary DNA (cDNA) (Zyagen). A T7 RNA polymerase recognition site was added to the 3′ end of the PCR product. The PCR product was purified using a PCR purification kit (Qiagen). One microgram of DNA was used for in vitro transcription by using digoxigenin (DIG) RNA labelling mix (Roche) and T7 RNA polymerase. After DNase I treatment for 30 min at 37 °C, the RNA probe was purified using probeQuant G-50 Columns (GE Healthcare). Sections (20 μm) were pre-treated with proteinase K (0.1 μg ml−1), acetylated, dehydrated through ethanol (50, 70, 95, and 100%), and air dried. Pre-treated sections were then incubated for 16–20 h at 60 °C, in a hybridization buffer containing sense or anti-sense riboprobes. After the sections were hybridized, they were treated with RNase A (20 μg ml−1) for 30 min at 37 °C and then washed four times in decreasing salinity (from 2× to 0.1× standard saline citrate buffer) and a 30 min wash at 68 °C. Sections were incubated with 3% hydrogen peroxide in PBS for 1 h and washed using PBS. After incubation in the blocking buffer for 1 h (TNB buffer, Perkin Elmer), sections were incubated with anti-DIG-POD antibody (1:500, Roche) in TNB buffer for 2 h. TSA-plus-Fluorescein reagent was used to visualize the signal. For GAD-FISH, anti-DIG-AP antibody (1:500, Roche) and Fast Red TR/Naphthol AS-MX (F4523, Sigma-Aldrich) were used to visualize the signal. After washing the sections in PBS, they were incubated with blocking buffer for 2 h followed by incubation with anti–GFP antibody overnight, and finally incubated with a secondary antibody as described above. To examine the overlap between each peptide marker and GAD, we used CCK-, CRH-, TAC1-, and PDYN-Cre mice injected with AAV-EF1α-DIO-ChR2–eYFP or AAV-EF1α-DIO–eYFP. These procedures are related to the results in Extended Data Figs 2 and 8. For analysis of rabies-tracing data, consecutive 60 μm coronal sections were collected and stained using Hoechst. Slides were scanned using Nanozoomer (Hamamatsu). GFP+ input neurons were counted from the forebrain to the posterior brainstem except sections adjacent to the injection sites (1 mm from the injection site), and grouped into ten regions based on Allen Mouse Brain Atlas (http://mouse.brain-map.org/static/atlas) using anatomical landmarks in the sections visualized by Hoechst staining and autofluorescence. We normalized the number of neurons in each region by the total number of input neurons in the entire brain. These procedures are related to the results in Extended Data Fig. 7. Consecutive 60 μm coronal sections were collected and stained using Hoechst. Slides were scanned using a Nanozoomer (Hamamatsu). All images were acquired using identical settings and were analysed using ImageJ as previously described15. Images were background subtracted (rolling ball radius of 50 pixels), thresholded, and pixels above this threshold were interpreted as positive signals. The mGFP- or eYFP-labelled axon arborization signal was measured for each region and averaged across the five sections. These procedures are related to the results in Extended Data Fig. 7. We adapted a previously described procedure to perform TRAP experiment39. Mice were euthanized at 12:00 to 14:00 and the POA was rapidly dissected on ice with a dissection buffer (1× HBSS, 2.5 mM HEPES (pH 7.4), 4 mM NaHCO , 35 mM glucose, 100 μg ml−1 cycloheximide). Brains from six mice were then pooled, homogenized in the homogenization buffer (10 mM HEPES (pH 7.4), 150 mM KCl, 5 mM MgCl , 100 nM calyculin A, 2 mM DTT, 100 U ml−1 RNasin, 100 μg ml−1 cycloheximide and protease). Homogenates were transferred to a microcentrifuge tube and clarified at 2,000g for 10 min at 4 °C. The supernatant was transferred to a new tube, and 70 μl of 10% NP40 and 70 μl of 1,2-diheptanoyl-sn-glycero-3-phosphocholine (DHPC, 300 mM) per millilitre of supernatant were added. This solution was mixed and then clarified at 17,000g for 10 min at 4 °C. The resulting high-speed supernatant was transferred to a new tube. This supernatant served as the input. A small amount (25 μl) was added to a new tube containing 350 μl of buffer RLT for future input RNA purification. Immunoprecipitation was performed with an anti-Flag antibody loaded beads. The beads were washed four times using 0.15 M KCl Wash buffer (10 mM HEPES (pH 7.4), 350 mM KCl, 5 mM MgCl , 2 mM DTT, 1% NP40, 100 U ml−1 RNasin, and 100 μg ml−1 cycloheximide). After the final wash, the RNA was eluted by addition of buffer RLT (350 μl) to the beads on ice, the beads removed by a magnet, and the RNA purified using the RNeasy Micro Kit (Qiagen) and analysed using an Agilent 2100 Bioanalyzer. cDNA libraries for RNA-seq were prepared with Ovation RNA-Seq System V2 and Ovation Ultralow Library Systems (NuGen), and analysed on an Illumina HiSeq 2500. Gene classification shown in Supplementary Table 1 was performed using PANTHER (http://pantherdb.org/)40. These procedures are related to the results in Fig. 4 and Extended Data Fig. 8. We adapted a previously described procedure to isolate fluorescently labelled neurons from the mouse brain41, 42, 43. Individual adult male mice (postnatal day 56 ± 3) were anaesthetized in an isoflurane chamber, decapitated, and the brain was immediately removed and submerged in fresh ice-cold artificial cerebrospinal fluid (ACSF) containing 126 mM NaCl, 20 mM NaHCO , 20 mM dextrose, 3 mM KCl, 1.25 mM NaH PO , 2 mM CaCl , 2 mM MgCl , 50 μM DL-AP5 sodium salt, 20 μM DNQX, and 0.1 μM tetrodotoxin, bubbled with a carbogen gas (95% O and 5% CO ). The brain was sectioned on a vibratome (Leica VT1000S) on ice, and each slice (300–400 μm) was immediately transferred to an ACSF bath at room temperature. After the brain slicing was complete (not more than 15 min), individual slices of interest were transferred to a small Petri dish containing bubbled ACSF at room temperature. The POA was microdissected under a fluorescence dissecting microscope, and the slices before and after dissection were imaged to examine the location of the microdissected tissue and confirm its location. The dissected tissue pieces were transferred to a microcentrifuge tube and treated with 1 mg ml−1 pronase (Sigma, P6911-1G) in carbogen-bubbled ACSF for 70 min at room temperature without mixing in a closed tube. After incubation, with the tissue pieces sitting at the bottom of the tube, the pronase solution was pipetted out of the tube and exchanged with cold ACSF containing 1% fetal bovine serum. The tissue pieces were dissociated into single cells by gentle trituration through Pasteur pipettes with polished tips of 600, 300, and 150 μm diameter. Single cells were isolated by fluorescence-activated cell sorting into individual wells of 96-well plates or 8-well PCR strips containing 2.275 μl of Dilution Buffer (SMARTer Ultra Low RNA Kit for Illumina Sequencing, Clontech 634936), 0.125 μl RNase inhibitor (SMARTer kit), and 0.1 μl of 1:1,000,000 diluted RNA spike-in RNAs (ERCC RNA Spike-In Mix 1, Life Technologies 4456740). Sorting was performed on a BD FACSAriaII SORP using a 130 μm nozzle, a sheath pressure of 10 p.s.i., and in the single-cell sorting mode. To exclude dead cells, DAPI (DAPI*2HCl, Life Technologies D1306) was added to the single-cell suspension to the final concentration of 2 ng ml−1. Sorted cells were frozen immediately on dry ice and stored at −80 °C. We used the SMARTer kit described above to reverse transcribe single-cell RNA and amplify the cDNA for 19 PCR cycles. To stabilize the RNA after quickly thawing the cells on ice, we immediately added to each sample an additional 0.125 μl of RNase inhibitor mixed with SMART CDS Primer II A. All steps downstream were performed according to the manufacturer’s instructions. cDNA concentration was quantified using Agilent Bioanalyzer High Sensitivity DNA chips. For most samples, 1 ng of amplified cDNA was used as input to make sequencing libraries with a Nextera XT DNA kit (Illumina FC-131-1096). Individual libraries were quantified using Agilent Bioanalyzer DNA 7500 chips. To assess sample quality and adjust the concentrations of libraries for multiplexing on HiSeq, all libraries were sequenced first on Illumina MiSeq to obtain approximately 100,000 reads per library, and then on Illumina HiSeq 2000 or 2500 to generate 100 base pair reads. These procedures are related to the results in Fig. 4 and Extended Data Fig. 8. Since both TRAP and single-cell RNA-seq have technical limitations and are prone to false-positive and false-negative errors, we used the following strategy for identifying markers for POA sleep neurons. (1) To eliminate false-positive errors, the candidate markers with existing Cre lines were tested in optogenetic experiments, and cell types that did not promote sleep were eliminated (for example, GAL, which was found to be enriched in the TRAP experiment). (2) To reduce false-negative errors, we included markers identified by either method in our candidate list, rather than only those identified by both methods. This should have enhanced our chance of finding a useful marker, even if it were missed by one of the methods because of false-negative errors. Of course, this strategy could increase the probability for false-positive errors in our candidate list, but these errors were eliminated by the functional test in (1). To inhibit CCK, CRH, or TAC1 neurons, we injected CNO dissolved in 0.1 ml vehicle solution (PBS with 0.5% dimethyl sulfoxide (DMSO)) into CCK-, CRH- or TAC1-Cre mice expressing hM4Di in the POA, 20 min before the recording session. CNO was administered intraperitoneally at 2.5 mg per kg (body weight). Vehicle solution was injected for the control experiment. These procedures are related to the results in Extended Data Fig. 10. Slice recordings were made at postnatal days 42–50. AAV -EF1α-DIO-ChR2–eYFP (500 nl) was injected into the POA of GAD2-Cre mice, and recording was made 2–3 weeks after injection. Slice preparation was according to procedures described previously44. A mouse was deeply anaesthetized with 5% isoflurane. After decapitation, the brain was dissected rapidly and placed in ice-cold oxygenated HEPES-buffered ACSF (in mM: NaCl 92, KCl 2.5, NaH PO 1.2, NaHCO 30, HEPES 20, glucose 25, sodium ascorbate 5, thiourea 2, sodium pyruvate 3, MgSO ·7H O 10, CaCl ·2H O 0.5, and NAC 12, at pH 7.4, adjusted with 10 M NaOH), and coronal sections of the TMN were made with a vibratome (Leica). Slices (300 μm thick) were recovered in oxygenated NMDG–HEPES solution (in mM: NMDG 93, KCl 2.5, NaH PO 1.2, NaHCO 30, HEPES 20, glucose 25, sodium ascorbate 5, thiourea 2, sodium pyruvate 3, MgSO ·7H O 10, CaCl ·2H O 0.5, and NAC 12, at pH 7.4, adjusted with HCl) at 32 °C for 10 min and then maintained in an incubation chamber with oxygenated standard ACSF (in mM: NaCl 125, KCl 3, CaCl 2, MgSO 2, NaH PO 1.25, sodium ascorbate 1.3, sodium pyruvate 0.6, NaHCO 26, glucose 10, and NAC 10, at pH 7.4, adjusted by 10 M NaOH) at 25 °C for 1–4 h before recording. All chemicals were from Sigma. Whole-cell recordings were made at 30 °C in oxygenated solution (in mM: NaCl 125, KCl 4, CaCl 2, MgSO 1, NaH PO 1.25, sodium ascorbate 1.3, sodium pyruvate 0.6, NaHCO 26, and glucose 10, at pH 7.4). Inhibitory postsynaptic currents were recorded using a caesium-based internal solution (in mM: CsMeSO 125, CsCl 2, HEPES 10, EGTA 0.5, MgATP 4, Na GTP 0.3, sodium phosphocreatine 10, TEACl 5, QX-314 3.5, at pH 7.3, adjusted with CsOH, 290–300 mOsm) and isolated by clamping the membrane potential of the recorded neuron at the reversal potential of the excitatory synaptic currents. The resistance of the patch pipette was 3–5 MΩ. The cells were excluded if the series resistance exceeded 40 MΩ or varied by more than 20% during the recording period. To activate ChR2, we used a mercury arc lamp (Olympus) coupled to the epifluorescence light path and bandpass filtered at 450–490 nm (Semrock), gated by an electromagnetic shutter (Uniblitz). A blue light pulse (5 ms) was delivered through a 40 × 0.8 numerical aperture water immersion lens (Olympus) at a power of 1–2 mW. Data were recorded with a Multiclamp 700B amplifier (Axon instruments) filtered at 2 kHz and digitized with a Digidata 1440A (Axon instruments) at 4 kHz. Recordings were analysed using Clampfit (Axon instruments). These procedures are related to the results in Extended Data Fig. 2. At the end of each recording, cytoplasm was aspirated into the patch pipette, expelled into a PCR tube as described previously45. The single-cell reverse-transcription PCR (RT–PCR) protocol was designed to detect the presence of mRNAs coding for GAPDH, GAD1, VGLUT2, and HDC. First, reverse transcription and the first round of PCR amplification were performed with gene-specific multiplex primer using the SuperScript III One-Step RT–PCR kit (12574-018, Invitrogen) according to the manufacturer’s protocol. Second, nested PCR was performed using GoTaq Green Master Mix (M7121, Promega) with nested primers for each gene. Amplification products were visualized via electrophoresis using 2% agarose gel. Primers (5′>3′) for single-cell RT–PCR were as follows. GAPDH (sense/anti-sense): multiplex, ACTCCACTCACGGCAAATTC/CACATTGGGGGTAG GAACAC; nested, AGCTTGTCATCAACGGGAAG/GTCATGAGCCCTTC CACAAT; Final product 331 base pairs (bp). GAD1 (sense/anti-sense): multiplex, CACAGGTCACCCTCGATTTT/TCTATGCCGCTGAGTTTGTG; nested, TAGCTGGTGAATGGCTGACA/CTTGTAACGAGCAGCCATGA; final product 200 bp. VGLUT2 (sense/anti-sense): multiplex, GCCGCTACATCATAGCCATC/GCTCTCTCCAATGCTCTCCTC; nested, ACATGGTCAACAACAGCACTATC/ATAAGACACCAGAAGCCAGAACA; final product 506 bp. HDC (sense/anti-sense): multiplex, GGAGCCCTGTGAATACCGTG/TCCACTGAAGAGTGAGCCTGA; nested, CGTGAATACTACCGAGCTAGAGG/ACTCGTTCAATGTCCCCAAAG; final product 182 bp. These procedures are related to the results in Extended Data Fig. 2. Statistical analysis was performed using MATLAB, GraphPad Prism, or Python. The selection of statistical tests was based on reported previous studies. All statistical tests were two-sided. The 95% confidence intervals for brain state probabilities were calculated using a bootstrap procedure: for an experimental group of n mice, with mouse i comprising m trials, we repeatedly resampled the data by randomly drawing for each mouse m trials (random sampling with replacement). For each of the 10,000 iterations, we recalculated the mean probabilities for each brain state across the n mice. The lower and upper confidence intervals were then extracted from the distribution of the resampled mean values. To test whether a given brain state was significantly modulated by laser stimulation, we calculated for each bootstrap iteration the difference between the mean probabilities during laser stimulation and the preceding period of identical duration. The investigators were not blinded to allocation during experiments and outcome assessment. To determine the sample size for optogenetic and pharmacogenetic experiments, we first performed pilot experiments with two or three mice. Given the strength of the effect and the variance across this group, we then predicted the number of animals required to reach sufficient statistical power. To determine the sample size (number of units) for optrode recordings, we first recorded from two animals. Given the success rate of finding identified units and the homogeneity of units in the initial data set, we set a target sample size. For rabies-mediated retrograde tracing, histology, and slice recording experiments, the selection of the sample size was based on numbers reported in previous studies. For gene profiling experiments, sample size was not calculated a priori, and the selection of the sample size was based on previous studies. Otherwise, no statistical methods were used to predetermine sample size. The single-cell RNA-seq data have been deposited in the Gene Expression Omnibus under accession number GSE79108. All other data are available from the corresponding author upon reasonable request.
News Article | April 25, 2017
A game called Mozak is turning thousands of Internet users into “tracers” who help neuroscientists map out the tangled circuitry of brain cells. The citizen-science project was created by the University of Washington’s Center for Game Science in partnership with the Allen Institute for Brain Science. Mozak took a share of the spotlight at last October’s White House Science Fair, but the project is just now coming out of beta. In a news release, UW says results gleaned from the game have helped the Allen Institute’s researchers reconstruct neurons 3.6 times faster than previous methods. Guided by online tutorials, the game’s tracers… Read More
News Article | April 25, 2017
Mozak: a game that crowdsources the detailed mapping of brain-cells Mozak is a game where you score points for participating in the mass-scale, crowdsourced mapping of dendrites in scanned brains of humans, rodents, and other organisms. The tracing-tasks are repeatedly performed by different players, who emerge a consensus that corrects for individual errors. It's similar to re:CAPTCHA, which used login/authentication screens to refine machine-vision guesses at the ambiguous words. The game itself is surprisingly fun! It was created by Microsoft co-founder Paul Allen's "Allen Institute for Brain Science" at the University of Washington -- which is led by Zoran Popović, who created the Foldit game that made significant advances in understanding protein folding. The key to Mozak is to allow humans to do what they are good at, and computers to do what they are good at. In this case, humans exceedingly outperform computers with their ability to resolve and trace objects in 3 dimensions. Although computers can complete the reconstruction of robustly traced neurons, many neurons have delicate and highly branched structures that your eyes will distinguish far better. Your ability to do this will help us to build systems that work better at doing this, and can lead to focused discoveries in this space. Like other science games, computers can quickly simulate, verify, and potentially perform local optimizations on player designs — something that would be tedious and inefficient for humans to perform. By discovering new 3-dimensional neuronal arborization patterns, eventually matching them to firing and genetic patterns, and then simulating them in the game, we hope that you will come up with new ideas that can be applied to ongoing research in the field.
News Article | April 27, 2017
Putting a turbo engine into an old car gives it an entirely new life--suddenly it can go further, faster. That same idea is now being applied to neuroscience, with a software wrapper that can be used on existing neuron tracing algorithms to boost their ability to handle not just big, but enormous sets of data. The wrapper, called UltraTracer, is highlighted this month in Nature Methods. "In trying to uncover the diversity of neuron shapes, scale is a very large and increasingly pressing problem," says Hanchuan Peng, Ph.D., Associate Investigator at the Allen Institute for Brain Science. "We need to be able to compare tens of thousands of neuron shapes in order to really understand what they look like, and to use that information to parse individual cell 'types.'" Peng and his team designed UltraTracer to work with existing neuron tracing algorithms designed by scientists around the world, turbo-charging them to work faster and with larger datasets. In the paper, they describe applying UltraTracer to ten different base tracers and also to any other base tracers in the BigNeuron initiative (bigneuron.org), developed by different people and that used varying techniques to automatically detect the shapes of neurons in three-dimensional image stacks. The team was able to demonstrate UltraTracer's unique ability to supercharge existing software. Using the Allen Cell Types Database as a biological reference, the software first learned what neurons "should" look like. UltraTracer then made existing algorithms more efficient to handle bigger data sets, and combined several different algorithms in an organic way that made the most of each tracer's strengths. "With UltraTracer, we are giving new life to neuron tracing algorithms that already exist and making them even more powerful," says Peng. "We can now test how well these algorithms work at very large scales, and make them work better. This will be a crucial step in addressing fundamental questions about cell types in the brain."
News Article | April 7, 2017
The Allen Institute for Brain Science has released additional data and computer models of cell activity for inclusion in the Allen Cell Types Database: a publicly available tool for researchers to explore and understand the building blocks of the brain. "Comprehensive coverage of hundreds to thousands of cells will be crucial for scientists who want to explore the diversity of nerve cells in the brain, and provides a base from which we can parse cells into meaningful types," says Lydia Ng, Ph.D., Senior Director of Technology at the Allen Institute for Brain Science. "This release is one more step in building a fundamental framework to help make advancements in neuroscience." Models serve as a critical link between observed data and theories about how cells work, enabling scientists to understand the mechanisms that give rise to neuron function. Two types of models have been added and updated as part of this release. The first set are models that reduce the complexity of neurons and use cell measurements to "predict" the activity and function of those cells, which are now available for 633 neurons in the database. Additionally, more sophisticated neuronal models based on cell shape, morphology and subcellular components are now available for hundreds of neurons via an interactive web browser. The Allen Cell Types Database contains detailed descriptive features gathered from individual neurons in the mouse brain, including location, electrical activity and shape. For this release, electrophysiological recordings from an additional 130 cells from the cortex have been added to the database. The Allen Cell Types Database (celltypes.brain-map.org) is a fundamental resource of the Allen Institute's ten-year plan to understand how activity in the brain leads to perception, decision-making and action. Understanding cell types--the brain's building blocks--is critical to making sense of both how the healthy brain functions and what goes wrong in diseases such as autism, Alzheimer's and Parkinson's. Additional updates to Allen Brain Atlas resources are planned for June and October of 2017.
News Article | April 27, 2017
Gamers can now significantly speed up the reconstructing of brain cells’ intricate architecture—a fundamental task in 21st century brain science. Mozak enables citizen scientists to produce complete, 3D reconstructions of neurons from different regions of the brain in animals and people. Figuring out the different shapes of nerve cells is a fundamental first step in analyzing how they assemble into the vast circuits that make up our brain. Since Mozak launched in November, the novice players—roughly 200 of them a day—and neuroscientists from the Allen Institute for Brain Science have been able to reconstruct neurons 3.6 times faster than previous methods. The game, developed at the University of Washington Center for Game Science, provides a framework to greatly increase the number of people who can tackle this core task in neuroscience. The players have also outperformed computers at tracing the complicated shapes of neurons. With minimal oversight, they can produce reconstructions that are 70 to 90 percent complete, compared to roughly 10 to 20 percent for the most effective computer-generated reconstructions. “New technologies have allowed us to create three-dimensional images of individual neurons, but our ability to catalog these brain cells, map their structure, and understand the relationships between them has been shockingly slow,” says Center for Game Science director Zoran Popović, a professor at the Paul G. Allen School of Computer Science & Engineering. “There’s a big bottleneck in processing and analyzing the data coming in, which is where the Mozak community is making a big impact.” Mozak has also let Allen Institute researchers shift away from tools that require complicated training and extensive expert input, without sacrificing quality. “Mozak is a great opportunity for us to work with citizen neuroscientists to answer questions about the diversity of cell types that exist in the brain and help us reach our scientific goals much quicker,” says Staci Sorensen, a morphology researcher at the Allen Institute for Brain Science. “It’s really exciting that regular people out in the world can, in a short period of time, be taught how to reconstruct neurons on the same level as experts who have been doing this a long time.” New players can also get real-time feedback from expert neuroscientists, a unique feature that allows Mozak players to acquire world-class expertise much faster. Mozak also requires general consensus among multiple players about a neuron’s shape, which allows for unprecedented levels of accuracy. By providing reconstructions that are confirmed not just by one or two scientists but by a collection of trained players working independently, Mozak in effect can provide neuroscience with the first validated, gold-standard reconstructions. The creation of Mozak got its inspiration partly from the growing needs for analyzing data from global projects like the BRAIN initiative, according to neuroscientist Jane Roskams, who served on the BRAIN advisory working group. “We still don’t know the answer to simple questions—like how many different types of neurons exist in a single brain—but understanding their physical form is a critical part of this. Mapping and analyzing neurons will help us understand how their structure relates to brain function, both in healthy brains, and as hallmarks of disease,” says Roskams. Mozak shows players a magnified volumetric image of a neuron—a key type of brain cell that transmits information throughout the nervous system—and asks the viewer to trace or draw its visible branches, which can also appear as disconnected dots in more challenging areas of data. Many neurons have delicate and highly branched structures that human eyes can distinguish far better than computers can alone, researchers say. People also tend to be much better at inferring the likely detailed structure from faint and sometimes discontinuous data. Yet computers are better at performing tedious tasks that can take humans a long time and are faster at reconstructing from clear and continuous data. Mozak brings together people and computing in a new way, to solve this enormous problem together. “This is not a story about people beating computers because people are using subsets of these computational tools,” Popović says. “This is about leveraging the things that computers can do well and the things that people do really well when they’re trained, forming a symbiotic superpower capable of solving unsolved challenges in science.” Until recently, neuroscience labs were doing well to trace about one neuron a week, says Stephen Smith, senior investigator for the Allen Institute for Brain Science. Mozak can speed that progress dramatically, but fully understanding neural networks will require tracing many thousands of neurons, Smith says. In the future, Mozak players may provide data that enable artificial intelligence and computer-vision tools to become smarter and more effective at reconstructing neurons on a much larger scale. “To make the kind of progress we wish to make in understanding the functioning of the human brain, we will eventually need to fully automate this task,” Smith says. “But to further develop our artificial intelligence and computer vision tools, we need a vast amount of traced and annotated data of the kind that Mozak’s citizen scientists will be producing. They will provide the ground truth.”