Ultrathin needle can deliver drugs directly to the brain

MIT researchers have devised a miniaturized system that can deliver tiny quantities of medicine to brain regions as small as 1 cubic millimeter. This type of targeted dosing could make it possible to treat diseases that affect very specific brain circuits, without interfering with the normal function of the rest of the brain, the researchers say.

Using this device, which consists of several tubes contained within a needle about as thin as a human hair, the researchers can deliver one or more drugs deep within the brain, with very precise control over how much drug is given and where it goes. In a study of rats, they found that they could deliver targeted doses of a drug that affects the animals’ motor function.

“We can infuse very small amounts of multiple drugs compared to what we can do intravenously or orally, and also manipulate behavioral changes through drug infusion,” says Canan Dagdeviren, the LG Electronics Career Development Assistant Professor of Media Arts and Sciences and the lead author of the paper, which appears in the Jan. 24 issue of Science Translational Medicine.

“We believe this tiny microfabricated device could have tremendous impact in understanding brain diseases, as well as providing new ways of delivering biopharmaceuticals and performing biosensing in the brain,” says Robert Langer, the David H. Koch Institute Professor at MIT and one of the paper’s senior authors.

Michael Cima, the David H. Koch Professor of Engineering in the Department of Materials Science and Engineering and a member of MIT’s Koch Institute for Integrative Cancer Research, is also a senior author of the paper.

Targeted action

Drugs used to treat brain disorders often interact with brain chemicals called neurotransmitters or the cell receptors that interact with neurotransmitters. Examples include l-dopa, a dopamine precursor used to treat Parkinson’s disease, and Prozac, used to boost serotonin levels in patients with depression. However, these drugs can have side effects because they act throughout the brain.

“One of the problems with central nervous system drugs is that they’re not specific, and if you’re taking them orally they go everywhere. The only way we can limit the exposure is to just deliver to a cubic millimeter of the brain, and in order to do that, you have to have extremely small cannulas,” Cima says.

The MIT team set out to develop a miniaturized cannula (a thin tube used to deliver medicine) that could target very small areas. Using microfabrication techniques, the researchers constructed tubes with diameters of about 30 micrometers and lengths up to 10 centimeters. These tubes are contained within a stainless steel needle with a diameter of about 150 microns. “The device is very stable and robust, and you can place it anywhere that you are interested,” Dagdeviren says.

The researchers connected the cannulas to small pumps that can be implanted under the skin. Using these pumps, the researchers showed that they could deliver tiny doses (hundreds of nanoliters) into the brains of rats. In one experiment, they delivered a drug called muscimol to a brain region called the substantia nigra, which is located deep within the brain and helps to control movement.

Previous studies have shown that muscimol induces symptoms similar to those seen in Parkinson’s disease. The researchers were able to generate those effects, which include stimulating the rats to continually turn in a clockwise direction, using their miniaturized delivery needle. They also showed that they could halt the Parkinsonian behavior by delivering a dose of saline through a different channel, to wash the drug away.

“Since the device can be customizable, in the future we can have different channels for different chemicals, or for light, to target tumors or neurological disorders such as Parkinson’s disease or Alzheimer’s,” Dagdeviren says.

This device could also make it easier to deliver potential new treatments for behavioral neurological disorders such as addiction or obsessive compulsive disorder, which may be caused by specific disruptions in how different parts of the brain communicate with each other.

“Even if scientists and clinicians can identify a therapeutic molecule to treat neural disorders, there remains the formidable problem of how to delivery the therapy to the right cells — those most affected in the disorder. Because the brain is so structurally complex, new accurate ways to deliver drugs or related therapeutic agents locally are urgently needed,” says Ann Graybiel, an MIT Institute Professor and a member of MIT’s McGovern Institute for Brain Research, who is also an author of the paper.

Measuring drug response

The researchers also showed that they could incorporate an electrode into the tip of the cannula, which can be used to monitor how neurons’ electrical activity changes after drug treatment. They are now working on adapting the device so it can also be used to measure chemical or mechanical changes that occur in the brain following drug treatment.

The cannulas can be fabricated in nearly any length or thickness, making it possible to adapt them for use in brains of different sizes, including the human brain, the researchers say.

“This study provides proof-of-concept experiments, in large animal models, that a small, miniaturized device can be safely implanted in the brain and provide miniaturized control of the electrical activity and function of single neurons or small groups of neurons. The impact of this could be significant in focal diseases of the brain, such as Parkinson’s disease,” says Antonio Chiocca, neurosurgeon-in-chief and chairman of the Department of Neurosurgery at Brigham and Women’s Hospital, who was not involved in the research.

The research was funded by the National Institutes of Health and the National Institute of Biomedical Imaging and Bioengineering.

de MIT News http://ift.tt/2rERXQc

How some facial malformations arise

About 1 in 750 babies born in the United States has some kind of craniofacial malformation, accounting for about one-third of all birth defects.

Many of these craniofacial disorders arise from mutations of “housekeeping” genes, so called because they are required for basic functions such as building proteins or copying DNA. All cells in the body require these housekeeping genes, so scientists have long wondered why these mutations would produce defects specifically in facial tissues.

Researchers at MIT and Stanford University have now discovered how one such mutation leads to the facial malformations seen in Treacher-Collins Syndrome, a disorder that affects between 1 in 25,000 and 1 in 50,000 babies and produces underdeveloped facial bones, especially in the jaw and cheek.

The team found that embryonic cells that form the face are more sensitive to the mutation because they more readily activate a pathway that induces cell death in response to stress. This pathway is mediated by a protein called p53. The new findings mark the first time that scientists have determined how mutations in housekeeping genes can have tissue-specific effects during embryonic development.

“We were able to narrow down, at the molecular level, how issues with general regulators that are used to make ribosomes in all cells lead to defects in specific cell types,” says Eliezer Calo, an MIT assistant professor of biology and the lead author of the study.

Joanna Wysocka, a professor of chemical and systems biology at Stanford University, is the senior author of the study, which appears in the Jan. 24 online edition of Nature.

From mutation to disease

Treacher-Collins Syndrome is caused by mutations in genes that code for proteins required for the assembly and function of polymerases. These proteins, known as TCOF1, POLR1C, and POLR1D, are responsible for transcribing genes that make up cell organelles called ribosomes. Ribosomes are critical to all cells.

“The question we were trying to understand is, how is it that when all cells in the body need ribosomes to function, mutations in components that are required for making the ribosomes lead to craniofacial disorders? In these conditions, you would expect that all the cell types of the body would be equally affected, but that’s not the case,” Calo says.

During embryonic development, these mutations specifically affect a type of embryonic cells known as cranial neural crest cells, which form the face. The researchers already knew that the mutations disrupt the formation of ribosomes, but they didn’t know exactly how this happens. To investigate that process, the researchers engineered larvae of zebrafish and of an aquatic frog known as Xenopus to express proteins harboring those mutations.

Their experiments revealed that the mutations lead to impairment in the function of an enzyme called DDX21. When DDX21 dissociates from DNA, the genes that encode ribosomal proteins do not get transcribed, so ribosomes are missing key components and can’t function normally. However, this DDX21 loss only appears to happen in cells that are highly sensitive to p53 activation, including cranial neural crest cells. These cells then undergo programmed cell death, which leads to the facial malformations seen in Treacher-Collins Syndrome, Calo says.

Other embryonic cells, including other types of neural crest cells, which form nerves and other parts of the body such as connective tissue, are not affected by the loss of DDX21.

Role of DNA damage

The researchers also found that mutations of POLR1C and POLR1D also cause damage to stretches of DNA that encode some of the RNA molecules that make up ribosomes. The amount of DNA damage correlated closely with the severity of malformations seen in individual larvae, and mutations in POLR1C led to far more DNA damage than mutations in POLR1D. The researchers believe these differences in DNA damage may explain why the severity of Treacher-Collins Syndrome can vary widely among individuals.

Calo’s lab is now studying why affected cells experience greater levels of DNA damage in those particular sequences. The researchers are also looking for compounds that could potentially prevent craniofacial defects by making the cranial neural crest cells more resistant to p53-induced cell death. Such interventions could have a big impact but would have to be targeted very early in embryonic development, as the cranial neural crest cells begin forming the tissue layers that will become the face at about three weeks of development in human embryos.

The research was funded by the National Institutes of Health, Howard Hughes Medical Institute, and March of Dimes Foundation.

de MIT News http://ift.tt/2rxHLsY

New type of virus found in the ocean

A type of virus that dominates water samples taken from the world’s oceans has long escaped analysis because it has characteristics that standard tests can’t detect. However, researchers at MIT and the Albert Einstein College of Medicine have now managed to isolate and study representatives of these elusive viruses, which provide a key missing link in virus evolution and play an important role in regulating bacterial populations, as a new study reports.

Viruses are the main predators of bacteria, and the findings suggest that the current view of bacterial virus diversity has a major blind spot. These conclusions have emerged through detailed analysis of marine samples led by MIT postdoc Kathryn Kauffman, professor of civil and environmental engineering Martin Polz, professor Libusha Kelly of Albert Einstein College of Medicine, and nine others. The results are being reported this week in the journal Nature.

The newly identified viruses lack the “tail” found on most catalogued and sequenced bacterial viruses, and have several other unusual properties that have led to their being missed by previous studies. To honor that fact, the researchers named this new group the Autolykiviridae — after a character from Greek mythology who was storied for being difficult to catch. And, unlike typical viruses that prey on just one or two types of bacteria, these tailless varieties can infect dozens of different types, often of different species, underscoring their ecological relevance.

This research “opens new avenues for furthering our understanding of the roles of viruses in the ocean,” says Jed Fuhrman, the McCulloch-Crosby Chair of Marine Biology at the University of Southern California, who was not involved in this work. “In a practical sense, it also shows how we need to alter some commonly used methods in order to capture these kinds of viruses for various studies,” he says. “I’d say it is an important advance in the field.”

Current environmental models of virus-bacteria interactions are based on the well-studied tailed viruses, Kauffman explains, so they may be missing important aspects of the interactions taking place in nature.

“We already knew that viruses are very important there,” Kauffman says, referring to the surface ocean, where the researchers’ samples were drawn, and where about 10 million viruses are found in every milliliter of water. Polz says that while “most of the viruses studied in labs have tails, most of those in the ocean don’t.” So the team decided to study one subset of tailless viruses, which infects a group of bacteria called Vibrio. After extensive tests, they found “that some of these were infecting unusually large numbers of hosts,” he says.

After sequencing representatives of the Autolykiviridae, the researchers found “their genomes were quite different from other viruses,” Polz says. For one thing, their genomes are very short: about 10,000 bases, compared to the typical 40,000-50,000 for tailed viruses. “When we found that, we were surprised,” he says.

With the new sequence information, the researchers were able to comb through databases and found that such viruses exist in many places. The research also showed that these viruses tend to be underrepresented in databases because of the ways samples are typically handled in labs. The methods the team developed to obtain these viruses from environmental samples could help researchers avoid such losses of information in the future. In addition, Kauffman says, typically the way researchers test for viral activity is by infecting bacteria with the viral sample and then checking the samples a day later to look for signs that patches of the bacteria have been killed off. But these particular nontailed viruses often act more slowly, and the killed-off regions don’t show up until several days have passed — so their presence was never noticed in most studies.

The new group of viruses may especially be widespread. “We don’t think it’s ocean-specific at all,” Polz says. For example, the viruses may even be prevalent in the human biome, and they may play roles in major biogeochemical cycles, he says, such as the cycling of carbon.

Another important aspect of theses findings is that the Autolykiviridae were shown to be  members of an ancient viral lineage that is defined by specific types of capsids, the protein shell encasing the viral DNA. Though this lineage is known to be very diverse in animals and protists — and includes viruses such as the adenoviruses that infect humans, and the giant viruses that infect algae — very few viruses of this kind have been found to infect bacteria.

“This work substantially changes the existing ideas on the composition of the ocean virome by showing that the content of small, tailless viruses … is comparable to that of the tailed viruses … that are currently thought to dominate the virosphere,” says Eugene V. Koonin, a senior investigator at the National Institutes of Health, who was not involved in this research. “This work is important also for understanding the evolution of the virus world because it shows that viruses related to the most common viruses of eukaryotes (such as adenoviruses, poxviruses, and others), at least in terms of the capsid structure, are much wider-spread in prokaryotes than previously suspected.”

Koonin adds, “I further wonder whether the viruses reported here might only represent the tip of the proverbial iceberg, because capsid proteins can be highly diverged in sequence so that many are missed even in sensitive database searches. The findings are also of practical importance because the tailless viruses appear to play a major ecological role in the ocean, being responsible for a substantial fraction of bacteria-killing.”

The work was supported by the National Science Foundation and the Woods Hole Oceanographic Institution’s Ocean Ventures Fund.

de MIT News http://ift.tt/2Gdwsta

Construction begins on a new student residence on Vassar Street

Working together in a process that is now familiar at MIT, a collaborative team of architects, students, and staff from the Division of Student Life and Campus Construction has developed a next-generation design for the new student residence on Vassar Street.

The new building, recently approved by the Cambridge Planning Board, will be constructed on the site of the West Garage parking facility and is expected to be open by the fall of 2020. Site-enabling activities began in September, and the demolition of the West Garage begins this month.

New student housing and housing renewal are current top priorities for MIT. The Institute’s commitment is reflected in additional projects such as the planned construction of a new graduate residence in Kendall Square and the current renovation of New House. The new Vassar Street residence hall will add 450 undergraduate beds and 12 Graduate Resident Tutor (GRT) apartments to MIT’s housing inventory. The building will also address a range of other student needs expressed through the collective design process.

Purposeful community engagement

During the project design phase, members of the MIT community — including the New Residences Working Group, comprised of students and staff from the Division of Student Life and Campus Construction — were active participants in the exploration of different design approaches and ideas. Central to this process was the identification of specific goals for the residence.

For the MIT participants, priorities included a design that would foster small, close-knit living communities and include a mix of first-year and upper-class students, a variety of community-building spaces, and a community kitchen available for use by students. For the architects, a key goal was a design that would be as unique and bold as other MIT campus architecture yet would fit its surrounding context and fully support the student communities living within.

“MIT is working to provide an on-campus housing experience that enhances students’ learning and personal development,” says Suzy M. Nelson, vice president and dean for student life. “The new Vassar Street residence design is the culmination of a process that began in the summer of 2016 to describe the ideal MIT undergraduate residential experience. That includes smaller clusters of students within the larger residence, ample flexible shared space, and food and dining facilities suitable for a wide range of student needs.” Throughout that process, notes Nelson, students have been deeply engaged and vocal about needs and expectations for the new residence. “Their input has been critical all the way along — from the architectural principles to the building design — because they have the best understanding of what works in the existing communities and how those features enhance the student experience.”

According to Michael Maltzan of Michael Maltzan Architecture (lead architects on the project), the Vassar Street project design process involved positive, lively debate and enabled the team to collectively seek balance by suggesting and testing various concepts.

“For example, we used the length of the site to real advantage,” he explains, “inside and out. The design will help create real and perceived social activity along the entire length.” Maltzan is also a lecturer in MIT's Department of Architecture.

A design that guides and enables

The building’s main entrance will be centered in its first floor and marked by a soaring blue wall on a grand scale. Surrounded by a mostly color-neutral environment of concrete and brick, this embellished wall is intended to draw the eye and invite social convergence.

Inside, the blue color will continue to thread its way through the building, highlighting the paths people will use as they move around. Student rooms will be arranged in clusters that mix single and double rooms with shared community spaces and a GRT apartment.  The building’s design is based in part on the “critical paths” students will take to reach their rooms within these clusters. At the same time, the design also encourages engagement among different clusters by making it easy for one to connect with others via well-placed stairways and gathering points.

On the ground floor, the residence will offer a 225-seat residential dining facility that is open to members of the community and includes a kitchen where students may also cook for themselves. The ground floor will also house an inviting avenue of common spaces such as study lounges, a private courtyard, and a makerspace area.

The ground floor design incorporates a variety of elements intended to foster social engagement. At the entrance, a courtyard with trees and a shaded seating area will welcome residents and community members. Much of the interior ground floor space will be visible or partly visible from the outside through glass walls and ribbons of glazing. Even the private courtyard will yield a sense of connection to the street through greenery-twined screened walls.

Sustainability: Aiming high

As is the norm for MIT’s design and construction initiatives, the project team considered and prioritized sustainable solutions at every level. With LEED Gold certification as its target, the building’s design incorporates a range of efficiency strategies that are now standard at MIT, including high-performance heating and cooling systems, efficient lighting and appliances, and stormwater management. The design also responds to MIT’s sustainability goals by establishing more than 275 new bicycle parking spaces for residents of the new building.

But to go one step further, the Vassar Street project team is incorporating sustainable construction techniques inspired by the Passive House standard that focuses on the energy efficiency of buildings. For example, the team expects the majority of the building’s exterior to be constructed using a panelized exterior system, where the panels are prefabricated and inspected in a factory prior to installation. The impacts of this technique include reducing the energy needed to heat or cool the building, based on exterior panels that allow very little air infiltration and reduce thermal bridging.

This construction process may help the residence hall serve as a test case for MIT as it evaluates sustainable options for other campus construction projects going forward.

“With this project, MIT continues to invest in a vibrant residential life experience for undergraduate students,” notes Richard Amster, director of campus construction. “This building will present the opportunity to enhance campus life, not only as a new, more sustainable facility but as a building that will enable us to continue renewing our existing housing stock.”

Serving its community

As envisioned by its collaborative design team, the Vassar Street residence will serve MIT in many other ways as well.

It will serve as a reflection of the neighborhood’s character and history, from its industrial-inspired north-facing walls that incorporate the scale and textures of the railway corridor to its south-facing courtyards and green spaces that provide connective pathways to its interior. It will serve the Institute at large with a new community gathering space at its western end, new benches and lighting along Vassar Street, and a welcoming new dining facility. Most important, it will serve its residents by providing vital student spaces close to the center of campus — spaces designed to foster inclusiveness and build community.

“I hope the building is an integral part [of students’ lives],” explains Michael Maltzan. “Then I think the building will be doing what architecture does at its very best, which is to be a productive, supportive, provocative armature for life in its most real way.”

Current construction activities

Construction activities at the Vassar Street site to date have included site-enabling activities such as establishing the necessary utilities for the new residence. The next step — the demolition of the West Garage — is expected to begin this month and be completed by April. At that point, construction of the residence foundation will begin, including the installation of precast concrete piles. Construction of the residence itself is expected to begin this fall.

During the construction, safety fencing will enclose the site along its perimeter, and new pedestrian crosswalks will redirect foot traffic to the south side of Vassar Street along the length of the construction site. In addition, a new shared bike lane will be created along the westbound side of Vassar Street (to be shared by bicycles and vehicles). The eastbound pedestrian and cycling lanes on the athletic field side of Vassar will remain unchanged. Two-way traffic will be maintained along Vassar Street, with a traffic detail in place as needed to ensure safety and access.

The project team is planning a series of communications to keep the community in the loop throughout the construction process. In addition to notifications that are being emailed and posted online, the team is also conducting information sessions with site abutters at major milestones along the way.

For more information about the construction project, please contact Senior Project Manager TJ Fanning. Online updates will also be available on the construction updates page on the Department of Facilities website. Members of the community can subscribe to email notifications regarding construction activities on the updates page as well.

de MIT News http://ift.tt/2DtDUP8

Designing vehicle-sharing networks

The proliferation of smartphones, vehicle-sharing apps, and traffic sensors has amounted to a wealth of data that can be used to provide insight for increasing the efficiency and sustainability of transportation networks.

Such data is particularly valuable to graduate students like Tianli Zhou, a PhD candidate in the Interdepartmental Program in Transportation in the Department of Civil and Environmental Engineering, who uses the information to design vehicle-sharing services.

“Car sharing became more popular in the last decade, so a lot of data has accumulated over the years,” Zhou says. “So the main questions are how do you help the practitioners of car-sharing services, and also the city planners, to design a better car-sharing system?"

With a background in industrial engineering, Zhou didn’t work on transportation systems until he was a junior at Tsinghua University in Beijing. There, Zhou worked closely with Professor Hai Jiang SM ’04, PhD ’06 to create an offline transportation itinerary planning app. Zhou received the Award for Exceptional Performance in Student Research Training Program at Tsinghua University for the app, and the app won second prize in the 2012 AutoNavi China Location Based Service Challenge. Zhou credits this experience with introducing him to the field of transportation.

“I think transportation is one of the most important topics in future urban contexts,” Zhou says. “Traffic congestion and the resulting air pollution are huge issues in many cities worldwide and I want to do something to mitigate this problem.”

For his master’s thesis at MIT — completed with Chancellor and Ford Professor of Engineering Cynthia Barnhart and Carolina Osorio, an associate professor of civil and environmental engineering — Zhou studied data from Hubway, the Boston area’s bicycle-sharing system, to see how bike sharing could be used to supplement the public transportation network and attract more individuals to use multiple modes of transportation for their trips or commutes.

Now, Zhou is working with Osorio and fellow graduate student Evan Fields, a PhD candidate in MIT’s Operations Research Center, to study data from the car-sharing company Zipcar. The project is funded by Ford.

For the project, Zipcar provided the researchers with “high resolution,” or extremely detailed, reservation data from Boston over a two-year period. Among the data was anonymized information on the location of preferred rental vehicles, reservation times, and the times users picked up and returned the vehicles.

Using this information, Zhou and Fields infer demand for vehicles and develop algorithms to inform best practices for car-sharing services and to make such services more convenient for users.

“Some researchers make assumptions about this type of data. We make a lot fewer assumptions and use the high resolution data to inform our work,” Zhou says. “We call this a ‘data-driven method,’ because we can use this data to make direct, evidence-based suggestions.”

Fields considers this data-driven method a highlight of the research project.

“This is a really fun project to work on because of the data we have from Zipcar; it’s so rich and complete. We have all of the reservation data from Boston for a two-year period, so we can see everything,” Fields says. “We can ask all types of questions like, ‘Do people like to use Zipcar on the weekends?’ or ‘Do people want long trips?’ We can look in the data and see the answers, and we know the data we get back is the truth. It is so rare to be able to write a query and see what happens for real.”

By creating and using a smart sampling strategy on the historical car reservation data provided by Zipcar, a simulator proposed by Fields can model the operation of the two-way car-sharing system. Such simulation can effectively replicate the real-life Zipcar fleet utilization rate and is used to infer true demand for the car-sharing service.

While most simulation methods proposed by previous studies do not scale-up to address car-sharing network design problems for large cities, Zhou proposed a new algorithm, with Fields’ high resolution data-driven simulator embedded, that allows the team to look at the greater Boston area. This algorithm produces suggestions based on the inferred demand, such as where Zipcar should locate its cars for the upcoming month.

“A large-scale analysis allows us to identify synergies with other mobility services provided throughout the city,” Osorio says. “In particular, we are currently investigating how car-sharing services can complement public transportation services to improve transportation accessibility across the city.”

The suggestions from the algorithm also have potential to both increase revenue for vehicle-sharing services and to make them more conveniently located for individual use. For example, by placing a certain number of cars in a specific area, with the preferred vehicles, and thus meeting user demand, individuals may be more likely to utilize these services, Zhou says.

“Helping Zipcar achieve their goals is helpful for everybody, particularly if Zipcar can help fill in gaps in the accessibility of a city, such as places where the T [subway system] doesn’t go,” Fields says. “If we can suggest where to locate the vehicles and simultaneously increase profit, that’s good for Zipcar. It keeps them around and incentivizes operations in Boston, but also provides transportation to the city.”

Zhou and Fields have submitted their Boston findings and algorithms to an academic journal, and have recently begun to apply the algorithm to similar Zipcar data from Manhattan.

Both researchers are advised by Osorio, whose research group's projects include traffic optimization, autonomous mobility, and vehicle sharing. Previous work in Osorio’s lab has focused on the sustainability benefits of optimizing traffic light monitoring; a 2015 study found that changing the timing of stoplights in urban areas could reduce greenhouse gas emissions.

In recent years, the group has developed models and algorithms to enable high-resolution mobility data, such as that of the car sharing project, to be used to optimize the design and the operations of mobility systems at the scale of full cities and metropolitan regions.

“Vehicle sharing is a way to enhance the sustainability of our transportation system. People don’t have to have their own vehicles, including both bikes and cars. Also, in the current car-sharing industry, there tend to be more compact cars, so it’s better for the environment,” Zhou says. “These aspects of my project makes me feel that I can have an impact on society, and that’s what interested me in this kind of research.”

de MIT News http://ift.tt/2DrMwFP

Lifting the veil on "valence," brain study reveals roots of desire and dislike

The amygdala is a tiny hub of emotions where in 2016 a team led by MIT neuroscientist Kay Tye found specific populations of neurons that assign good or bad feelings, or “valence,” to experience. Learning to associate pleasure with a tasty food, or aversion to a foul-tasting one, is a primal function and key to survival.

In a new study in Cell Reports, Tye’s team at the Picower Institute for Learning and Memory returns to the amygdala for an unprecedentedly deep dive into its inner workings. Focusing on a particular section called the basolateral amygdala, the researchers show how valence-processing circuitry is organized and how key neurons in those circuits interact with others. What they reveal is a region with distinct but diverse and dynamic neighborhoods where valence is sorted out by both connecting with other brain regions and sparking cross-talk within the basolateral amygdala itself.

“Perturbations of emotional valence processing is at the core of many mental health disorders,” says Tye, associate professor of neuroscience at the Picower Institute of Learning and Memory and the Department of Brain and Cognitive Sciences. “Anxiety and addiction, for example, may be an imbalance or a misassignment of positive or negative valence with different stimuli.”

Despite the importance of valence assignment in both healthy behavior and psychiatric disorders, neuroscientists don’t know how the process really works. The new study therefore sought to expose how the neurons and circuits are laid out and how they interact.

Bitter, sweet

To conduct the study, lead author Anna Beyeler, a former postdoc in Tye’s lab and currently a faculty member at the University of Bordeaux in France, led the group in training mice to associate appealing sucrose drops with one tone and bitter quinine drops with another. They recorded the response of different neurons in the basolateral amygdala when the tones were played to see which ones were associated with the conditioned learned valence of the different tastes. They labeled those key neurons associated with valence encoding and engineered them to become responsive to pulses of light. When the researchers then activated them, they recorded the electrical activity not only of those neurons but also of many of their neighbors to see what influence their activity had in local circuits.

They also found, labeled, and made similar measurements among neurons that became active on the occasion that a mouse actually licked the bitter quinine. With this additional step, they could measure not only the neural activity associated with the learned valence of the bitter taste but also that associated with the innate reaction to the actual experience.

Later in the lab, they used tracing technologies to highlight three different kinds of neurons more fully, visualizing them in distinct colors depending on which other region they projected their tendrilous axons to connect with. Neurons that project to a region called the nucleus accumbens are predominantly associated with positive valence, and those that connect to the central amygdala are mainly associated with negative valence. They found that neurons uniquely activated by the unconditioned experience of actually tasting the quinine tended to project to the ventral hippocampus.

In all, the team mapped over 1,600 neurons.

To observe the three-dimensional configuration of these distinct neuron populations, the researchers turned the surrounding brain tissues clear using a technique called CLARITY, invented by Kwanghun Chung, assistant professor of chemical engineering and neuroscience and a colleague in the Picower Institute.

Neighborhoods without fences

Beyeler, Tye, and their co-authors were able to make several novel observations about the inner workings of the basolateral amygdala’s valence circuitry.

One finding was that the different functional populations of neurons tended to cluster together in neighborhoods, or “hotspots.” For example, picturing the almond-shaped amygdala as standing upright on its fat bottom, the neurons projecting to the central amygdala tended to cluster toward the point at the top and then on the right toward the bottom. Meanwhile the neurons that projected to the nucleus accumbens tended to run down the middle, and the ones that projected to the hippocampus were clustered toward the bottom on the opposite side from the central amygdala projectors.

Despite these trends, the researchers also noted that the neighborhoods were hardly monolithic. Instead, neurons of different types frequently intermingled creating a diversity where the predominant neuron type was never far from at least some representatives of the other types.

Meanwhile, their electrical activity data revealed that the different types exerted different degrees of influence over their neighbors. For example, neurons projecting to the central amygdala, in keeping with their association with negative valence, had a very strong inhibitory effect on neighbors, while nucleus accumbens projectors had a smaller influence that was more balanced between excitation and inhibition.

Tye speculates that the intermingling of neurons of different types, including their propensity to influence each other with their activity, may provide a way for competing circuits to engage in cross-talk.

“Perhaps the intermingling that there is might facilitate the ability of these neurons to influence each other,” says Tye.

Notably, Tye’s research has indicated the projections the different cell types make appear immutable, but the influence those cells have over each other is flexible. The basolateral amygdala may therefore be arranged to both assign valence and negotiate it, for instance in those situations when a mouse spies some desirable cheese, but that mean cat is also nearby.

“This helps us understand how form might give rise to function,” says Tye.

In addition to Beyeler and Tye, the paper’s other authors are Chia-Jung Chang, Margaux Silvestre, Clementine Leveque, Praneeth Namburi, and Craig Wildes.

Several funding sources, including the JPB Foundation, Whitehall Foundation, Klingenstein Foundation, Alfred P. Sloan Foundation, New York Stem Cell Foundation, and the National Institutes of Health provided support for the study.

de MIT News http://ift.tt/2G9Hdg1

Bots that talk more like people

Before coming to MIT, Jeff Orkin SM ’07, PhD ’13 spent a decade building advanced, critically acclaimed artificial intelligence (AI) for video games.

While working on F.E.A.R., a survival-horror first-person shooter game, he developed AI that gave computer-controlled characters an unprecedented range of actions. Today, more than 10 years later, many video game enthusiasts still consider the game’s AI unmatched, even by modern standards.

But for Orkin, the game’s development inspired a new line of interest. “A big focus of that game was getting squads of enemy characters to work together as a team and communicate constantly,” Orkin says. “That got me really into thinking about, ‘How do you get machines to converse like humans?’”

Following the game’s release in 2005, Orkin enrolled in the MIT Media Lab, where he spent the next eight years tackling that challenge. Now, through his startup Giant Otter Technologies, he’s using his well-honed AI skills to help chatbots expertly navigate tricky human conversations.

Giant Otter’s platform uses AI algorithms and crowdsourced annotators to build a natural-language database, compiled “bottom-up,” from archived sales and customer support transcripts. Chatbots draw on this robust database to better understand and respond, in real time, to fluctuating, nuanced, and sometimes vague language.

“[The platform] was inspired by the way episodic memory works in the human mind: We understand each other by drawing from past experiences in context,” says Orkin, now Giant Otter’s CEO. “The platform leverages archived data to understand everything said in real time and uses that to make suggestions about what a bot should say next.”

The startup is currently piloting the platform with e-commerce and telecommunication companies, pharmaceutical firms, and other large enterprises. Clients can use the platform as a “brain” to power a Giant Otter chatbot or use the platform’s conversation-authoring tools to power chatbots on third-party platforms, such as Amazon’s Lex or IBM’s Watson. The platform automates both text and voice conversations.

Benefits come in the form of cost savings. Major companies can spend billions of dollars on sales and customer support services; automating even a fraction of that work can save millions of dollars, Orkin says. Consumers, of course, will benefit from smarter bots that can more quickly and easily resolve their issues.

Human-machine collaboration

In conversation, people tend to express the same intent with different words, potentially over several sentences, and in various word orders. Unlike other chatbot-building platforms, Giant Otter uses “human-machine collaboration,” Orkin says, “to learn authentic variation in the way people express different thoughts, and to do it bottom-up from real examples.”

Giant Otter’s algorithms comb through anywhere from 50 to 100 transcripts from sales and customer support conversations, identifying language variations of the same intent, such as “How can I help you?” and “What’s your concern?” and “How may I assist you?” These are called “utterances.” All utterances are mixed around into chunks of test scripts for people to judge for accuracy online.

Consider a script for a sales call, where a salesperson is selling a product while the prospect is pushing for a discount. Giant Otter’s algorithms match and substitute one utterance in one script with a similar one from another script — such as swapping “I may be able to offer a discount” with “I’ll see if I can reach your price point.” That version is uploaded to Mechanical Turk or another crowdsourcing platform, where people will vote a “yes” or “no” if the substituted sentence makes sense.

In another human task, people break conversations into “events.” Giant Otter will lay out conversations horizontally and people will label different sections of the conversation. A salesperson saying, “Hello, thanks for contacting us,” for instance, may be labeled as “call opening.” Other section labels include “clarifying order,” “verifying customer information,” “proposing resolution,” and “resolving issue.”

“Between these two tasks, we learn a lot about the structure over how conversations unfold,” Orkin says. “Conversations break down into events, events break down into utterances, and utterances break down into many different examples of saying the same thing with different words.”

This builds a robust language database for chatbots to recognize anywhere from a few to more than 100 different ways to express the same sentiment — including fairly abstract variations. This is important, Orkin says, as today’s chatbots are built top-down, by a human manually plugging in various utterances. But someone seeking an order status update could say, for instance, “My order hasn’t come, and I checked my account, and it said to contact customer support.”

“Nowhere does the person even say ‘status.’ If I was creating content for a bot by hand, there’s no way I would have thought of that,” Orkin says. The platform continues to learn and evolve after chatbots are deployed.

Path to chatbots

Giant Otter’s origin goes back to 2005, when Orkin joined the Cognitive Machines Group led by Deb Roy, an associate professor of media arts and sciences. Roy had just initiated the Human Speechome Project, his effort to gather data on how humans develop language by video and audio recording his newborn for three years.

Branching off from the project, Orkin developed The Restaurant Game, an online game that paired people online to have natural, text-based discussions as a customer and a waitress at a restaurant. “We hoped to get maybe 100 people to play. But not too long after [the game launched], we had data from 16,000 people,” Orkin says.

Many people would order food, pay bills, and talk about the menu. Others, however, were more unorthodox, asking the waitress on a date, stealing the cash register, or stacking cakes up to climb onto the roof. All of that data was valuable. “Whatever they did, we had all that data of natural conversations between players,” Orkin says. Soon, he hired people around the world through crowd-labor platforms, such as Mechanical Turk and Upwork, to ascribe context to the game’s transcripts.

The game turned into a platform to collect and curate dialog to make AI conversations more natural. In 2013, Orkin met Geoff Marietta, a Harvard Graduate School of Education student studying how virtual worlds could facilitate learning and improve relationships. Using Orkin’s platform, they developed a game called SchoolLife, where players assume roles of a bullying victim and a bystander. Players interacted with AI-controlled characters to come up with a solution to student conflict.

To commercialize the game, the two co-founded Giant Otter in 2013. For early support, Orkin turned to the MIT Venture Mentoring Service. Introductions through VMS have since led to a valuable partnership and potential pilot customers.

SchoolLife earned a Small Business Innovation Research grant from the National Science Foundation and was used in numerous schools in the region. But the budgeting cycles of school districts made it difficult for a startup to thrive in the education sector. Moreover, simulating bullying posed an issue, Orkin says. “With our platform, you need to capture data reflecting how people naturally converse. With bullying, we did come up with ways to record conversations from people role-playing in simulated scenarios, but it wasn’t authentic data,” Orkin says.

A third co-founder, Dan Tomaschko MBA ’15, soon recognized the potential to impact the corporate world and guided a pivot to a training tool, called Coach Otto, that trained employees to deal with sensitive scenarios in the workplace. But last fall the startup realized the niche market of chatbot development for customer support was more profitable. (Companies, however, can still use the platform to practice phone conversations for sales and support training.)

Currently, Giant Otter is working on better integrating its conversation-authoring tools with third-party chatbot platforms. It’s also developing its own automated customer support chatbot, powered by any company’s call transcripts. “It’s taken us years to realize where the most value is, but we’re focused on … [having] the right assembly line to crank enterprise phone call and live chat transcript data through our platform and turn it into something that can automate chatbot conversations,” Orkin says.

de MIT News http://ift.tt/2rwYzQO