More accurate computer models of molecules could help predict useful properties for everything from new drugs to better batteries. But simulating the behavior of the atoms and electrons they consist of means calculating huge numbers of possibilities, so even powerful computers use approximations.
Abhinav Kandala is solving this problem by using quantum computers to simulate molecules. In 2017 he simulated three-atom beryllium hydride, the largest molecule modeled on a quantum computer to date. This was a crucial step that laid the groundwork for precise simulations of larger molecules, which could lead to the discovery of new medicines and materials.
Quantum computers are made of qubits—the physical elements that encode information the way bits do in a conventional computer. Because qubits are governed by quantum mechanics, they could model other particles subject to its rules, like atoms and electrons, more easily than conventional computers. Kandala, who works for IBM Research in New York, says this makes simulating molecules one of the technology’s “killer applications.”
Since 2017 he’s made an even more fundamental contribution. Because quantum states are fragile, quantum computers are error-prone, and compensating for this requires large numbers of qubits. But today’s devices consist of only tens of qubits—not enough to create a fault-tolerant quantum computer. Kandala has demonstrated a way to harness errors to boost accuracy without increasing the number of qubits. His experiments allow him to identify trends that could be used to extrapolate what should be observed in the absence of errors, an advance that could speed practical applications of near-term quantum computers.
After Jason Buenrostro graduated from Santa Clara University with a degree in biology and engineering, he went to work in a lab at Stanford, overseeing an $800,000 gene sequencing machine.
He wanted to understand the effects of the genetic mutations his machine detected. But many of the mutated genes he found were considered junk because they didn’t direct the production of proteins. So in his graduate research at Stanford, he pivoted to developing methods for measuring these underexplored regions.
DNA is essentially identical from cell to cell, but a kidney cell differs from a brain cell in the activity of those genes. Regions of DNA need to be tightly wound to fit inside the nucleus, and only open DNA regions can be active.
Buenrostro and his colleagues developed a tool called ATAC-seq to measure these open regions of the genome, many of which don’t make proteins but regulate genetic activity. “I didn’t realize how useful [a tool] it would be for people. It kind of exploded,” says Buenrostro, noting that ATAC-seq now has its own Wikipedia page.
More recently, Buenrostro has further developed the technology to identify open DNA at the level of a single cell. With this tool, researchers can determine which genes are active in single cells, studying how these cells sometimes develop into cells of new types, and how some functions go awry in disease.
Buenrostro wants to use these methods to learn new basic information about the differences between healthy and diseased cells and to use this information to engineer new behaviors into cells as they develop and mature.
Buenrostro, who now oversees a lab of 10, says, “I want to understand cell fate decisions to ultimately be able to engineer cells to do whatever I would like them to”—for instance, fighting cancer.
In 20 years, antibiotic drug resistance is projected to kill more people than cancer. That’s why Silvia Caballero feels such urgency to develop new approaches to controlling bacterial infections.
She was among the first to discover that certain organisms among the trillions that inhabit the human gut can help the body fight back when antibiotic-resistant bacteria begin to take hold.
While working in a lab at Memorial Sloan-Kettering Cancer Center in New York, Caballero developed lab mice that mimic intestinal colonization by vancomycin--resistant enterococcus and carbapenem-resistant enterobacteriaceae, also known as superbugs. She used these models together with bioinformatic tools to identify species of microbes that could clear the mouse gut of multi-drug- resistant bacteria, in this way destroying the main reservoir for infection.
Now working for Vedanta Biosciences in Cambridge, Massachusetts, Caballero is trying to do the same for people, identifying bacteria that can effectively control three potentially lethal bacterial strains often found in hospitals and nursing homes.
She played a key role in the creation of the world’s largest library of human gut bacteria and led a campaign to test thousands of species for their ability to kill those three menacing organisms. Her work led to the identification of a bacterial cocktail derived from human gut flora that can control all three types of bacteria. Vedanta’s goal is to begin clinical studies with this drug candidate in 2021.
Light-emitting diodes (LEDs) are used in a plethora of products ranging from smartphone and TV screens to traffic lights, but they’re expensive to make. In addition, the sweet spot between the highly efficient conversion of electricity to light and the ability to shine brightly has been difficult to reach.
Dawei Di co-invented new LED materials and devices that can generate light from electricity at maximum efficiency even when they need to reach high brightness. What’s more, they can be manufactured using cheaper, simpler, and less energy-intensive processes.
Typical LED production lines require high-temperature processes or depositing light-emitting materials on a solid surface in a vacuum, and thus they use lots of energy. Di’s materials are cheaper because they can be made from widely available substances, and they don’t need to be deposited at high temperature or in a vacuum. Instead, they’re dissolved in a liquid and then coated onto a solid surface.
A number of companies are already testing production lines with Di’s methods. Although these lines won’t be replacing standard production facilities immediately, he believes they will become increasingly common. “The industry is heading that way,” says Di, who’s both a faculty member at Zhejiang University and a visiting researcher at the University of Cambridge.
Modern gene sequencing machines are very fast, reading through the DNA of a peanut, eggplant, or armadillo in two days. But what they spit out are billions of disorganized fragments of DNA code. Olga Dudchenko has helped to make the next step—pasting those bits together in the right order, to reveal the actual genome—faster and cheaper.
Dudchenko uses Hi-C, a technique originally developed to study how chromosomes fold, to show which bits of DNA lie physically close to one another. Coupled with Dudchenko’s methods and algorithms, this makes assembling genomes easy.
In late 2018, Dudchenko and her colleagues shared the first results of DNA Zoo, including end-to-end chromosome sequences for more than 50 species, including the cheetah, red panda, and Brazilian porcupine. In a world of mounting extinction, these species’ DNA code may one day be all that’s left of them.
The job ahead is to characterize the genome of every species on earth. The DNA Zoo (where Dudchenko is referred to as “chief zookeeper”) releases new data every week. “The ability to [make] decisions in an informed fashion can mean the difference between survival and extinction of the species,” she says.
One of the most promising cancer therapies to emerge in recent years is CAR T-cell therapy. This genetically alters a patient’s white blood cells, or T cells, to target a specific protein, or antigen, found on the surface of cancer cells before releasing chemicals to kill them. The problem is that cancer cells often share antigens with cells of other types, so the therapy is currently limited to cancers of certain blood cells with unique antigens.
Marc Lajoie has invented a way to reprogram T cells so they can target combinations of antigens rather than just single ones, which should allow them to tackle a much wider range of cancers. “It’s the equivalent of putting a microchip into a cell,” he says. “We can install these new programs and co-opt the cell to make the decisions that we want them to make.”
Lajoie and colleagues at the University of Washington developed switches made from proteins, which he then used as the basis of a series of logic gates capable of carrying out the same “and,” ”or,” and “not” operations that computer chips do.
Such gates can be tuned to react to different antigens, which allows T cells to target unique combinations of antigens, avoid antigens found on healthy cells, or target cancers that develop resistance due to antigen loss.
Lajoie has cofounded a startup called Lyell Immunopharma and works at the company’s Seattle office to develop more effective CAR T-cell therapies using his protein logic. But he says the same technology could help treat all kinds of diseases by rewiring how cells respond to their environment.
Ritu Raman’s robots are made out of both polymers and muscle tissue, and are capable of sensing their environment and recognizing temperature, pH, and mechanical pressure.
“I’m a mechanical engineer by training, and I’m honestly a little bored building with the materials we’ve been building with for the past thousand years. So I’m making robots and machines that use biological materials to move and walk around and sense their environment, and do more interesting things—like get stronger when they need to and heal when they get damaged.”
Raman has built 3D printers capable of patterning living cells and proteins, injecting those into a mold where the cells self-assemble into dense muscle tissue. The tissue is then transferred to a robotic skeleton. The robots, powered by living skeletal muscle, move in response to light or electricity.
Right now, they look a bit like inchworms, but that’s just the proof of concept. “Can we make new ‘biohybrid’ implants for drug delivery that adapt to your body better than purely synthetic implants could?” Raman says. “Can we release robots into a polluted water supply and have them walk toward a toxin and exude a chemical to neutralize that?”
Isaac Sesi built a gadget he believes can tackle one of the biggest risks faced by farmers across Africa: the contamination of grains following harvest.
Sesi’s product, GrainMate, allows famers and grain purchasers to affordably measure moisture levels of maize, rice, wheat, millet, sorghum, and other staples. It’s designed for a simple yet persistent problem: according to the UN Food and Agriculture Organization, more than 20% of sub--Saharan Africa’s cereal output is lost or wasted, often because grains aren’t dried sufficiently before they’re stored. Grain stored while moist can develop aflatoxins—contaminants produced by fungi that are harmful to humans and animals.
In Sesi’s native Ghana, individual farmers often sell their harvests to aggregators or animal feed producers; if one farmer’s crops are too moist they risk spoiling the entire batch. Although imported moisture detection devices are available, few farmers in Ghana can afford the nearly $400 price tag. “That might be half of what a farmer is making from his entire field” per harvest, Sesi says.
Sesi, who grew up without electricity or running water and often went to school hungry, spent much of his childhood tinkering with electronic devices. He learned by dissecting broken radios and other abandoned gadgets with the help of a book from his school library. He long sought a way to apply that passion to a field that could have a social impact—and in 2017, as a recent electrical engineering graduate, he got his chance. A United States Agency for International Development project operating in partnership with his school, the Kwame Nkrumah University of Science and Technology, had recently designed a grain-moisture meter for the local market. But it wanted to bring the cost down and find a way to produce the device in Ghana.
Sesi was their man: with the help of a small team, he streamlined the original device, redesigned its circuit board, built an accompanying mobile app, and found five Ghanaian subcontractors to make components that had previously been sourced from China. Sesi’s device sells for $80—less than one-fourth as much as existing alternatives. Sesi and his team are now developing a more efficient version of the meter and a second product to help farmers identify ideal soil inputs. They’re also raising funds to expand to the bigger markets in Kenya and Nigeria. Ultimately, Sesi believes he can help farmers across the continent cut wastage, minimize economic losses, and improve the safety of their products.
Brandon Sorbom has solved a fundamental problem that has made fusion reactors too expensive to build. By developing an electromagnetic system using high-temperature superconductors to insulate part of the fusion process, Sorbom’s breakthrough could make fusion reactor designs dramatically cheaper to build.
A fusion reactor that can deliver energy to the grid is more than a decade away at best. But developing such a reactor is a worthy goal because fusion has the potential to offer almost limitless zero-carbon energy, with low radioactive waste and safety risks.
One puzzle has stumped scientists for decades: how to maintain the 100 million-degree temperatures needed for fusion and do it cheaply enough to profitably produce energy. Powerful magnets can do the job by insulating the fuel at a reactor’s core. But until recently, not even the world’s best electromagnets were good enough.
So Sorbom and his team designed a better magnet from a superconductor called yttrium barium copper oxide. First as a student at MIT, and now as the chief scientist at startup Commonwealth Fusion Systems, Sorbom used this magnet as part of a fusion reactor design almost 100 times smaller than was previously thought possible. The reactor is so small, in fact, that Commonwealth Fusion is on track to build its first functional concept within the next decade.
Archana Venkataraman is using artificial intelligence to better map the human brain—and to develop entirely new ways to diagnose and treat neurological disorders.
Despite decades of research, we have only a basic understanding of disorders such as epilepsy, autism, Alzheimer’s, and schizophrenia, and thus a limited ability to treat them. Most therapies are administered on a trial and error basis, guided by a physician’s instinct. Many of them regularly fail.
Informed by data from existing imaging technologies—including the electroencephalogram, or EEG, and functional magnetic resonance imaging, or fMRI— Venkataraman develops mathematical models designed to unlock the “black box” of the brain’s function and provide the building blocks for treatments that are less invasive and far more precise. Her most groundbreaking work targets epilepsy, which affects more than 50 million people globally. Roughly 30% of epilepsy patients do not respond to medication and thus require surgery—which can only work after the seizure onset zone has been successfully isolated to a specific region of the brain.
Data-driven models that pinpoint seizure onset, Venkataraman believes, can limit invasive monitoring and improve surgical outcomes. She has developed a seizure-detection algorithm, which is being evaluated on clinical data from Johns Hopkins. This algorithm uses EEG data and employs methods of deep learning to track the time and location of seizure onset in patients’ brains.