New 3-D printing approach makes cell-scale lattice structures

A new way of making scaffolding for biological cultures could make it possible to grow cells that are highly uniform in shape and size, and potentially with certain functions. The new approach uses an extremely fine-scale form of 3-D printing, using an electric field to draw fibers one-tenth the width of a human hair.

The system was developed by Filippos Tourlomousis, a postdoc at MIT’s Center for Bits and Atoms, and six others at MIT and the Stevens Institute of Technology in New Jersey. The work is being reported today in the journal Microsystems and Nanoengineering.

Many functions of a cell can be influenced by its microenvironment, so a scaffold that allows precise control over that environment may open new possibilities for culturing cells with particular characteristics, for research or eventually even medical use.

While ordinary 3-D printing produces filaments as fine as 150 microns (millionths of a meter), Tourlomousis says, it’s possible to get fibers down to widths of 10 microns by adding a strong electric field between the nozzle extruding the fiber and the stage on which the structure is being printed. The technique is called melt electrowriting.

“If you take cells and put them on a conventional 3-D-printed surface, it’s like a 2-D surface to them,” he explains, because the cells themselves are so much smaller. But in a mesh-like structure printed using the electrowriting method, the structure is at the same size scale as the cells themselves, and so their sizes and shapes and the way they form adhesions to the material can be controlled by adjusting the porous microarchitecture of the printed lattice structure.

“By being able to print down to that scale, you produce a real 3-D environment for the cells,” Tourlomousis says.

He and the team then used confocal microscopy to observe the cells grown in various configurations of fine fibers, some random, some precisely arranged in meshes of different dimensions. The large number of resulting images were then analyzed and classified using artificial intelligence methods, to correlate the cell types and their variability with the kinds of microenvironment, with different spacings and arrangements of fibers, in which they were grown.

Cells form proteins known as focal adhesions at the places where they attach themselves to the structure. “Focal adhesions are the way the cell communicates with the external environment,” Tourlomousis says. “These proteins have measurable features across the cell body allowing us to do metrology. We quantify these features and use them to model and classify quite precisely individual cell shapes.”

For a given mesh-like structure, he says, “we show that cells acquire shapes that are directly coupled with the substrate’s architecture and with the melt electrowritten  substrates,” promoting a high degree of uniformity compared to nonwoven, randomly structured  substrates. Such uniform cell populations could potentially be useful in biomedical research, he says: “It is widely known that cell shape governs cell function and this work suggests a shape-driven pathway for engineering and quantifying cell responses with great precision,” and with great reproducibility.

He says that in recent work, he and his team have shown that certain type of stem cells  grown in such 3-D-printed meshes survived without losing their properties for much longer than those grown on a conventional two-dimensional substrate. Thus, there may be medical applications for such structures, perhaps as a way to grow large quantities of human cells with uniform properties that might be used for transplantation or to provide the material for building artificial organs, he says. The material being used for the printing is a polymer melt that has already been approved by the FDA.

The need for tighter control over cell function is a major roadblock for getting tissue engineering products to the clinic. Any steps to tighten specifications on the scaffold, and thereby also tighten the variance in cell phenotype, are much needed by this industry, Tourlomousis says.

The printing system might have other applications as well, Tourlomousis says. For example, it might be possible to print “metamaterials” — synthetic materials with layered or patterned structures that can produce exotic optical or electronic properties.

The team included Thrasyvoulos Karydis and Andreas Mershin at MIT, and Chao Jia, Hongjun Wang, Dilhan Kalyon, and Robert Chang at the Stevens Institute of Technology in Hoboken, New Jersey. The work was funded by the National Science Foundation.

de MIT News https://ift.tt/2TyhUJQ

Fine-tuning multiphysics problems

“Stretching myself radically to learn a new kind of physics or code is exactly what I want to do,” says Miriam Kreher. “It’s how I solve problems and find new ones.”

A second-year doctoral student in nuclear science and engineering, Kreher is finding just the kind of challenges she craves as a member of MIT’s Computational Reactor Physics Group (CRPG). Her task: helping to develop vastly improved software simulations of the complex interactions taking place inside nuclear reactors. 

“Some people focus on how neutrons move, and others look at how water flowing around the core affects temperature,” she explains. “But in nuclear reactors, these physics phenomena of neutron transport and fluid flow affect each other through complex feedback, and we need to understand both at the same time.” 

This tight coupling of physics phenomena has preoccupied nuclear engineering for some time. “Getting a more precise picture of these interactions would allow for finer-tuned operational margins in reactors,” notes Kreher, a Department of Energy Computational Science Graduate Fellow. More accurate simulations could help the current fleet of commercial reactors work at higher powers, and aid in designing the next generation of reactors.

However, modeling multiphysics problems is not simple. Even in steady-state, there are countless neutrons interacting with the fuel and coolant, depositing large amounts of energy that alters the temperature of everything inside the reactor. Add a time variable, or alter the position of the control rod, which determines the rate of fission reactions, and the modeling proves more difficult still. Current high-fidelity simulations are expensive, requiring weeks or longer to render.  But quite recently, scientists have begun to gain ground on these problems.

“Computers have now become powerful enough to address these multiphysics problems, permitting stable simulations in a shorter amount of time,” says Kreher. “This could be the basis for much less expensive modeling.”

Under the supervision of CRPG faculty leads Kord Smith and Benoit Forget, Kreher is developing computational tools that will yield high fidelity simulations with representative temperature and density conditions inside a reactor core. To tackle the complex multiphysics problems she confronts as she goes about this task, Kreher is taking a sequence of tough math and computation classes so she can test new modeling approaches.

“I want to help develop computational methods that will permit other researchers to simplify or speed up their simulations, so eventually they won’t need to depend on the world’s fastest computers,” she says. “I am excited to be part of something that could set the groundwork for science of the future.”

Kreher found her way into nuclear engineering early. Brought up in Morocco and France, she arrived at a magnet high school for science and technology in Marseille. There, Kreher became engaged by a class combining physics and English that included a survey on energy.

“I thought nuclear made the most sense, since it produced energy and cleaned up the atmosphere, so I decided I should contribute to the field somehow,” she recalls. She pursued engineering at the University of Pittsburgh — the city where her father grew up — concentrating in nuclear engineering. She found mentors who offered her research opportunities, and received her first taste of coding. “I give Pitt a lot of credit for that; I got my first internship at Bettis Atomic Power Laboratory because I knew MATLAB.”

Kreher also plunged into policy work, joining the Nuclear Engineering Student Delegation in 2014 for a week-long visit to Washington, where she learned about the intersection of politics and technology.

“If I had been drawn to fields other than math and physics, I might have become an advocate or lobbyist, because public perception of nuclear power is really important,” she says. To this day, Kreher says she likes to pop in on her Washington representative to discuss energy issues when her schedule permits.

She applied to the MIT Summer Research Program in 2015 to sharpen her resume for graduate school, and landed a research spot with Benoit Forget’s group.

“I was just an intern, but all the students helped me, and I eventually contributed to the group’s research,” says Kreher. “It felt fun and dynamic, a natural fit, and I decided on MIT for graduate school.”

Her work on multiphysics problems evolved quickly during meetings with Forget and Smith, who became her advisors. Summer research at several Department of Energy laboratories provided her opportunities to acquire additional coding techniques.

With her anticipated graduation in 2022, Kreher has years more of meticulous computation before her. “I keep the larger research vision in view, and while I struggle with coding issues from time to time I always get a rush when I solve them, which makes it feel worthwhile to go on to the next problem.”

For research breaks, she works up a sweat swing dancing with MIT’s Lindy Hop Society. “I have no background in dancing, but it makes me really happy, especially since I’m not one of those people who exercise,” she says. “It just gives me those natural endorphins from moving, plus it’s a social outlet for me.” And as MIT co-president of the student section of the American Nuclear Society, Kreher engages in the kind of outreach work on energy issues that remain important to her.

At the end of the doctoral road, a position at one of the national labs or perhaps a faculty position, beckons. In the meantime, MIT life is working out well. “Being here is very special, because I can problem solve with people, and they share things with me,” she says. “Culturally, socially, I’m very happy at MIT.”

She’s also a big fan of the 32-year-old MSRP, and of Institute efforts to make the science and engineering communities more inclusive.

de MIT News https://ift.tt/2HTtA8a

Exceptional individuals receive 2019 MIT Excellence Awards and Collier Medal

Eleven individuals and three teams were award recipients at the 2019 MIT Excellence Awards and Collier Medal ceremony on Friday, March 22.

Among the highest honors awarded to staff at the Institute, the Excellence Awards acknowledge our community’s extraordinary dedication to MIT’s goals, values, and mission, and recognize colleagues who excel in service to us all.

The Collier Medal honors the memory of Officer Sean Collier, who gave his life protecting and serving the MIT community nearly six years ago, and celebrates an individual or group whose actions demonstrate the importance of community.

The 2019 MIT Excellence Award recipients are:

• Mercedes Balcells-Camps, William H. Kindred, and the MIT Press Diversity and Inclusion Working Group Team in the category of Advancing Inclusion and Global Perspectives;
• Elizabeth DeRienzo and the International Students Office Team in the category of Bringing out the Best;
• Debby Carr in the category of Innovative Solutions;
• Chris Budny, Emily Gallagher, Shikha Sharma, and Claire Walsh in the category of Outstanding Contributor;
• Gerry O'Toole and Donyatta Small in the category of Serving the Client; and
• The Summit Farms Solar Power Purchase Agreement Team in the category of Sustaining MIT.

The 2019 Collier Medal recipient is Arman Rezaee, PhD candidate in the Department of Electrical Engineering and Computer Science in the School of Engineering.

Visit the MIT HR website for more information about the award categories, selection process, and recipients.

de MIT News https://ift.tt/2Op9mnF

New approach could boost energy capacity of lithium batteries

Researchers around the globe have been on a quest for batteries that pack a punch but are smaller and lighter than today’s versions, potentially enabling electric cars to travel further or portable electronics to run for longer without recharging. Now, researchers at MIT and in China say they’ve made a major advance in this area, with a new version of a key component for lithium batteries, the cathode.

The team describes their concept as a “hybrid” cathode, because it combines aspects of two different approaches that have been used before, one to increase the energy output per pound (gravimetric energy density), the other for the energy per liter (volumetric energy density). The synergistic combination, they say, produces a version that provides the benefits of both, and more.

The work is described today in the journal Nature Energy, in a paper by Ju Li, an MIT professor of nuclear science and engineering and of materials science and engineering; Weijiang Xue, an MIT postdoc; and 13 others.

Today’s lithium-ion batteries tend to use cathodes (one of the two electrodes in a battery) made of a transition metal oxide, but batteries with cathodes made of sulfur are considered a promising alternative to reduce weight. Today, the designers of lithium-sulfur batteries face a tradeoff.

The cathodes of such batteries are usually made in one of two ways, known as intercalation types or conversion types. Intercalation types, which use compounds such as lithium cobalt oxide, provide a high volumetric energy density — packing a lot of punch per volume because of their high densities. These cathodes can maintain their structure and dimensions while incorporating lithium atoms into their crystalline structure.

The other cathode approach, called the conversion type, uses sulfur that gets transformed structurally and is even temporarily dissolved in the electrolyte. “Theoretically, these [batteries] have very good gravimetric energy density,” Li says. “But the volumetric density is low,” partly because they tend to require a lot of extra materials, including an excess of electrolyte and carbon, used to provide conductivity.

In their new hybrid system, the researchers have managed to combine the two approaches into a new cathode that incorporates both a type of molybdenum sulfide called Chevrel-phase, and pure sulfur, which together appear to provide the best aspects of both. They used particles of the two materials and compressed them to make the solid cathode. “It is like the primer and TNT in an explosive, one fast-acting, and one with higher energy per weight,” Li says.

Among other advantages, the electrical conductivity of the combined material is relatively high, thus reducing the need for carbon and lowering the overall volume, Li says. Typical sulfur cathodes are made up of 20 to 30 percent carbon, he says, but the new version needs only 10 percent carbon.

The net effect of using the new material is substantial. Today’s commercial lithium-ion batteries can have energy densities of about 250 watt-hours per kilogram and 700 watt-hours per liter, whereas lithium-sulfur batteries top out at about 400 watt-hours per kilogram but only 400 watt-hours per liter. The new version, in its initial version that has not yet gone through an optimization process, can already reach more than 360 watt-hours per kilogram and 581 watt-hours per liter, Li says.  It can beat both lithium-ion and lithium-sulfur batteries in terms of the combination of these energy densities. 

With further work, he says, “we think we can get to 400 watt-hours per kilogram and 700 watt-hours per liter,” with that latter figure equaling that of lithium-ion. Already, the team has gone a step further than many laboratory experiments aimed at developing a large-scale battery prototype: Instead of testing small coin cells with capacities of only several milliamp-hours, they have produced a three-layer pouch cell (a standard subunit in batteries for products such as electric vehicles) with a capacity of more than 1,000 milliamp-hours. This is comparable to some commercial batteries, indicating that the new device does match its predicted characteristics.

So far, the new cell can’t quite live up to the longevity of lithium-ion batteries in terms of the number of charge-discharge cycles it can go through before losing too much power to be useful. But that limitation is “not the cathode’s problem”; it has to do with the overall cell design, and “we’re working on that,” Li says. Even in its present early form, he says, “this may be useful for some niche applications, like a drone with long range,” where both weight and volume matter more than longevity.

“I think this is a new arena for research,” Li says.

The work was supported by the Samsung Advanced institute of Technology, the National Key Technologies R&D Program of China, the National Science Foundation of China, and MIT’s Department of Materials Science and Engineering. The team also included professor Jing Kong and others at MIT, as well as researchers at the Chinese Academy of Sciences in Beijing, the Songshan Lake Materials Laboratory in Guangdong, China, the Samsung Advanced Institute of Technology America in Burlington, Massachusetts, and Tongji University in Shanghai.

de MIT News https://ift.tt/2YlIIAU

3 Questions: Why are student-athletes amateurs?

Debate about the unpaid status of NCAA athletes has surged in the last decade — and did so again last month when the best player in men’s college basketball, Zion Williamson, got injured in a high-profile game. Meanwhile, graduate student unionization drives frequently raise the same question: Aren’t some students also workers creating value for universities? And how did we come to regard student-athletes, say, as amateurs in the first place?

Jennifer Light, the Bern Dibner Professor in the History of Science and Technology and a professor of urban studies and planning, has just published an article in the Harvard Educational Review on the history of this idea that students are not part of the labor force. She places its origins in the 1890-1930 movement to expand public schooling, which promoted schools as alternatives to child labor and put them forth as “protected” places for young people to focus on future-oriented training. MIT News talked to Light about her research. This interview has been edited for length.

Q: How did you become interested in the topic of value-producing students, and the question of whether or not they’re fairly compensated?

A: Previously, I taught at Northwestern University, where I encountered many student-athletes because two of my classes covered some sports history. On more than one occasion, someone raised the question, “Why are we not getting compensated when we bring in so much money [for the university]?” Or I’d hear anecdotally about how a video game company came to scan their bodies for a game that was “cool,” but also not something they got paid for.

That got me thinking about unpaid labor. Of course, student-athletes receive scholarships, but those are quite limited when compared with the compensation they’d get playing outside of school. So I went looking for the origins of the idea that students are not part of the labor force. As it turns out, this idea dates to the emergence of mass schooling in the United States. As it also turns out, there is a long history of schools profiting from student activities — and not just sports.

Q: The “alternative history” of students you describe primarily occurs from about 1890 to 1930, with the movement to make public education available for everyone. What happened in this time period that was so important, in this regard?

A: Public education became popular, along with compulsory-schooling legislation, largely because of the industrial economy. The spread of schools was part of a national effort to train children for future industrial jobs and reduce child labor. And at some level, yes, when kids went to school, they were protected from going into factories or coal mines.

On the other hand, because so many public schools needed to get off the ground at the same time and local governments did not have adequate resources, educators assigned pupils to build and operate their schools: making desks and lockers, building playground equipment and gyms, keeping financial records, running the lunch room, everything from ordering supplies to cooking and serving the meal. Kids repaired school plumbing and heating systems, did health inspections, and tracked down truants.

This was celebrated as the cutting-edge curriculum, the "new education" for the industrial age. John Dewey said when you bring the school close to life, that motivates students’ learning. Because these things were done for educational purposes and no money exchanged hands, they were not considered "work." Of course, if kids did the same tasks in the “real world,” they would be paid. We still use this language of school versus the real world today. So this mindset originated in public schools and only later carried over into education for older adolescents, which was less common, particularly before World War II.

Q: How and when did the idea of the “protected” student make the leap from public schools to universities?

A: In the American mindset, until about 1930, you were a kid until you were between about 14 and 16, because in most places high school was not compulsory; education was compulsory until the eighth grade. And people were fighting to change this, but there were plenty of late teenagers in the work force.

Mass unemployment during the Great Depression was a catalyst for extending this protective period to older adolescents, through their early 20s. [The thinking was] that adults should be top priority for available jobs. So although increasingly specialized jobs were a contributing factor, the desire not to compete with adults was a major force behind the expansion of training for this age group — in high schools, community colleges, universities, and specialized programs such as those sponsored by the National Youth Administration. As with the curriculum for younger pupils, institutional maintenance was a feature of these programs. 

In recent years this sort of routine economic activity inside schools has declined but not disappeared. Today's controversies around student-athletes and teaching assistants stem in part from the century-old assumption that students by definition are cultivating their human capital and defering economic participation until they graduate to the “real world.” What I’m trying to show is, that’s always been a fantasy. Of course when students go to school, they get an education. But they could also be producing value for their institutions. 

de MIT News https://ift.tt/2WoYr0j

Developing tech for, and with, people with disabilities

Lora Brugnaro says to think of her like a Weeble toy that constantly wobbles then falls down. She has cerebral palsy, which severely impacts her balance, and for years she has used a walker to help her stay upright while moving around. Unfortunately, she has found that walkers available on the market are cheap, unstable, and prone to flipping on rough surfaces, leaving her sprawled out on the floor of an MBTA station or in the middle of the street. She had even started considering using a wheelchair to avoid such situations.

"I have felt for a very long time that the daily choice I made between safety and living with the freedom to move was an unnecessary choice predicated on poor design," she says.

Brugnaro was one of the co-designers at the sixth annual Assistive Technologies Hackathon, or ATHack. The event pairs teams of students, most but not all of whom study at MIT, with people from the Boston and Cambridge area living with disabilities.

Team Lora spent the hackathon working to create a more stable walker. Team member Zoe Levitt, walked with Brugnaro through a typical journey she takes to get from home to work and back again. Levitt observed Brugnaro's challenges and took videos to share with the rest of the team, supplementing Brugnaro's feedback.

"The collaboration was the most meaningful part of the event for me," Brugnaro says. "I’ve spoken to many people about my struggles; nobody ever offered solutions or suggested there might be any. It’s depressing and demeaning when companies ignore your feedback, and medical professionals seem satisfied with the standard options."

For her part, Levitt found Brugnaro's engaged collaboration instrumental in designing a product that worked; "Lora is super-awesome; she had a really clear idea of a problem she needed help with and was really eager to hear our ideas," she says. "Actually, after the hackathon … [Brugnaro] took the walker home, and on Monday I got this email with about 15 bullet points about what worked and what didn't work, and what could be improved!"

ATHack was founded in 2014 by MIT undergraduates Jaya Narain, Ishwarya Ananthabhotla, and Abigail Klein. They had previously been a part of tech projects that involved at the design stage the people who would ultimately be using the technology, and they had found the experience meaningful and valuable. They thought that other MIT students could benefit from a similar experience, and so created the ATHack to bring together community members living with disabilities and student designers.

"We wanted to create a way for a lot of people at MIT to become exposed to assistive technology through the co-design model," Narain says.

Unlike much of the research and development that takes place at MIT, the solutions created by the ATHack designers are often very straightforward and even simple. To Ananthabhotla, that simplicity is a marker of success.

"Unlike most other hackathons, this event is really not about building the best, 'coolest' tech," she explains. "It is about building a relationship with a community member and building the design skills needed to identify a need, understand a context, and ideate iteratively." 

Narain and Ananthabhotla organized MIT ATHack 2019 with Hosea Siu ’14, SM ’15, PhD ’18 and student volunteers Tareq El Dandachi, Imane Bouzit, Sally Beiruti, and Samuel Mendez. Together, they set up a Meet the Co-Designers dinner on Feb.11, where groups of students were paired up with co-designers on the bases of skills and interests.

The students had a few weeks to collaborate with their co-designer, brainstorm, and request specialized materials before the hackathon at the MIT Lincoln Laboratory Beaver Works Center on March 2. Teams were also allowed to start building before the hackathon, if they chose to do so.

Alex Rosenberg, another ATHack co-designer, uses a wheelchair and has limited arm strength and an inability to use his fingers.

"Having just become disabled, the most difficult thing is to find time for the problem solving of everyday life problems," Rosenberg says. "The hospitals, the therapists, and family and friends take care of the major things, but picking up a cup or throwing a ball to your kids are deemed by everybody else as not important enough to spend precious time figuring out how to make easier. The hackathon is so valuable because it makes space for that time."

Team Alex created a system to launch and catch a ball so that Alex could play with his two sons. They were particularly focused on making it simple enough for Alex to use it without the aid of another adult. The team came in second in the co-designer collaboration category, but more importantly, Alex has already been able to make use of the technology.

"The product we created at the hackathon as a team has already brought so much joy to my relationship with my sons," he says.

Sara Falcone was part of Team Reese. After first being introduced to him at the Meet the Co-Designers dinner, Falcone and her teammates at ATHack started designing a robust neck support brace. Reese, a teenager with cerebral palsy, has difficulty controlling his head motion, making it hard for him to use eye-tracking-based communication tools, watch TV, see the board during classes, and drive his chair. Team Reese created a pneumatic brace, which uses pressurized air to keep his head stable, to allow him and his family to control how much support he has at any given time.

"Reese was an awesome co-designer and a great person to meet; I learned a lot from him," Falcone says. "Since he is mostly nonverbal, I was initially concerned we wouldn't be able to make something that really worked for him, but he was able to fully communicate with few words. Reese has a contagious smile and laughed at all our failed first attempts, which really made us feel comfortable trying a bunch of different, goofy, often terrible ideas with him, and made for a fun day."

Winning teams were selected in four categories: usability, co-designer collaboration, technical innovation, and documentation. Team Reese came in first in the technical innovation category for their pneumatic neck brace. But Falcone isn't satisfied.

"I'm excited to keep working on our solution for Reese to make it a polished, durable tool for him," she says. "We ended the hackathon with a pretty solid direction, but the prototype was made in less than 12 hours, so there's a number of things we'd like to improve to make it better for Reese."

The winner in the usability category was Team Sara, who developed a portable bidet that would allow their co-designer to use the bathroom at work and in other public places. They published an Instructable on their product, so that other people can build one too.

Team Phil won the documentation category with a low-cost modular stander to aid people with limited lower body control. They published an entire website, with Ikea-like instructions on how to build their modular stander they named "PhilGood."

Finally, the co-designer collaboration category was won by Team Lora. Brugnaro used her improved walker on the way home from the hackathon, and has continued to use it every day since. She says she hopes to eventually bring the design to other people with mobility issues.

"I no longer feel like a second-class citizen with a crappily [designed], cheap rollator that breaks apart every six months," she says. "I finally have my own personal vehicle, with a brake system that actually brakes and a wheel system that provides 100 percent stability."

Cameron Taylor, one of the designers in Team Sara, sums up why ATHack is so important.

"Technologies are designed for those who fit standard parameters, leaving people who have unique physiologies technologically behind," he says. "From this economic perspective, having a unique physiology doesn’t make you disabled, it makes your technology disabled. Because the designers at ATHack do not come to the hack driven by economics, the hackathon breaks this cycle and allows technologies to be developed for people on an individual level."

Brugnaro puts it even more simply: "Without this hackathon, it would have been impossible to work with people who understood the science behind my problem and proved that there is a solution. Two weeks, four bright students, and one user changed my life."

de MIT News https://ift.tt/2HM548J

Model learns how individual amino acids determine protein function

A machine-learning model from MIT researchers computationally breaks down how segments of amino acid chains determine a protein’s function, which could help researchers design and test new proteins for drug development or biological research. 

Proteins are linear chains of amino acids, connected by peptide bonds, that fold into exceedingly complex three-dimensional structures, depending on the sequence and physical interactions within the chain. That structure, in turn, determines the protein’s biological function. Knowing a protein’s 3-D structure, therefore, is valuable for, say, predicting how proteins may respond to certain drugs.

However, despite decades of research and the development of multiple imaging techniques, we know only a very small fraction of possible protein structures — tens of thousands out of millions. Researchers are beginning to use machine-learning models to predict protein structures based on their amino acid sequences, which could enable the discovery of new protein structures. But this is challenging, as diverse amino acid sequences can form very similar structures. And there aren’t many structures on which to train the models.

In a paper being presented at the International Conference on Learning Representations in May, the MIT researchers develop a method for “learning” easily computable representations of each amino acid position in a protein sequence, initially using 3-D protein structure as a training guide. Researchers can then use those representations as inputs that help machine-learning models predict the functions of individual amino acid segments — without ever again needing any data on the protein’s structure.

In the future, the model could be used for improved protein engineering, by giving researchers a chance to better zero in on and modify specific amino acid segments. The model might even steer researchers away from protein structure prediction altogether.

“I want to marginalize structure,” says first author Tristan Bepler, a graduate student in the Computation and Biology group in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “We want to know what proteins do, and knowing structure is important for that. But can we predict the function of a protein given only its amino acid sequence? The motivation is to move away from specifically predicting structures, and move toward [finding] how amino acid sequences relate to function.”

Joining Bepler is co-author Bonnie Berger, the Simons Professor of Mathematics at MIT with a joint faculty position in the Department of Electrical Engineering and Computer Science, and head of the Computation and Biology group.

Learning from structure

Rather than predicting structure directly — as traditional models attempt — the researchers encoded predicted protein structural information directly into representations. To do so, they use known structural similarities of proteins to supervise their model, as the model learns the functions of specific amino acids.

They trained their model on about 22,000 proteins from the Structural Classification of Proteins (SCOP) database, which contains thousands of proteins organized into classes by similarities of structures and amino acid sequences. For each pair of proteins, they calculated a real similarity score, meaning how close they are in structure, based on their SCOP class.

The researchers then fed their model random pairs of protein structures and their amino acid sequences, which were converted into numerical representations called embeddings by an encoder. In natural language processing, embeddings are essentially tables of several hundred numbers combined in a way that corresponds to a letter or word in a sentence. The more similar two embeddings are, the more likely the letters or words will appear together in a sentence.

In the researchers’ work, each embedding in the pair contains information about how similar each amino acid sequence is to the other. The model aligns the two embeddings and calculates a similarity score to then predict how similar their 3-D structures will be. Then, the model compares its predicted similarity score with the real SCOP similarity score for their structure, and sends a feedback signal to the encoder.

Simultaneously, the model predicts a “contact map” for each embedding, which basically says how far away each amino acid is from all the others in the protein’s predicted 3-D structure — essentially, do they make contact or not? The model also compares its predicted contact map with the known contact map from SCOP, and sends a feedback signal to the encoder. This helps the model better learn where exactly amino acids fall in a protein’s structure, which further updates each amino acid’s function.

Basically, the researchers train their model by asking it to predict if paired sequence embeddings will or won’t share a similar SCOP protein structure. If the model’s predicted score is close to the real score, it knows it’s on the right track; if not, it adjusts.

Protein design

In the end, for one inputted amino acid chain, the model will produce one numerical representation, or embedding, for each amino acid position in a 3-D structure. Machine-learning models can then use those sequence embeddings to accurately predict each amino acid’s function based on its predicted 3-D structural “context” — its position and contact with other amino acids.

For instance, the researchers used the model to predict which segments, if any, pass through the cell membrane. Given only an amino acid sequence, the researchers’ model predicted all transmembrane and non-transmembrane segments more accurately than state-of-the-art models.

“The work by Bepler and Berger is a significant advance in representing the local structural properties of a protein sequence,” says Serafim Batzoglou, a professor of computer science at Stanford University. “The representation is learned using state-of-the-art deep learning methods, which have made major strides in protein structure prediction in systems such as RaptorX and AlphaFold. This work has ultimate application in human health and pharmacogenomics, as it facilitates detection of deleterious mutations that disrupt protein structures.”

Next, the researchers aim to apply the model to more prediction tasks, such as figuring out which sequence segments bind to small molecules, which is critical for drug development. They’re also working on using the model for protein design. Using their sequence embeddings, they can predict, say, at what color wavelengths a protein will fluoresce.

“Our model allows us to transfer information from known protein structures to sequences with unknown structure. Using our embeddings as features, we can better predict function and enable more efficient data-driven protein design,” Bepler says. “At a high level, that type of protein engineering is the goal.”

Berger adds: “Our machine learning models thus enable us to learn the ‘language’ of protein folding — one of the original ‘Holy Grail’ problems — from a relatively small number of known structures.”

de MIT News https://ift.tt/2TYwOOY