Lung Organoids: A Novel Way to Model COVID Infection

Calvin Kuo, MD, PhD, with Shannon Choi, MD, PhD, a student in the Kuo lab. Courtesy Steve Fisch

A year into the pandemic, we’ve all heard the stories. A patient is a little short of breath but appears to have a mild case of COVID-19. The next day, she deteriorates so rapidly that she’s rushed to intensive care, put on a ventilator, and hooked up to a dialysis machine to prevent kidney failure. Her overzealous immune system has gone rogue, attacking healthy cells instead of just fighting off the virus.

What triggers this devastating immune response, called a cytokine storm? Researchers are still struggling to identify the underlying processes that initiate a COVID infection and subsequent cytokine storm.

Biologists use advanced technologies and cell cultures in petri dishes to study severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the coronavirus strain responsible for COVID-19, identifying its key characteristics such as the famous crownlike spikes on their surfaces. But these short-lived cultures don’t act like real organs. And scientists are limited by their samples.

“When you analyze samples from patients, they’re often at the end stage of the disease, and many of the samples are from autopsy. You can’t understand the initiation process because the tissue is essentially destroyed,” says Calvin Kuo, MD, PhD, professor of hematology.

Understanding how the disease develops and testing potential treatments require better ways to model this coronavirus.

Miniature Organs in a Dish

Kuo’s laboratory develops organoids—three-dimensional miniature organs grown in a petri dish that mimic the shape, structure, and tissue organization of real organs.

Grown from human tissue samples using precisely defined ingredients, these organoids are little spheres of gel up to 1 millimeter in diameter. Healthy tissue samples are mechanically minced and enzyme digested to get to single cells, and then the organoids are grown from single stem cells. They last about six months, significantly longer than the few-weeks lifetime of traditional cell cultures.

Kuo initially developed organoids to study stem cell biology and model cancer. His team was the first to use organoids to convert normal tissues to cancer, as previously reported in Nature Medicine.

But he was passionate about using organoids to model infectious diseases. In 2015, he led a National Institute of Allergy and Infectious Diseases U19 research program, recently renewed for an additional five years, in collaboration with Stanford researchers Manuel Amieva, MD, professor of pediatrics and of microbiology and immunology; Harry Greenberg, MD, the Joseph D. Grant Professor in the Stanford University School of Medicine and professor of microbiology and immunology; Elizabeth Mellins, MD, professor of pediatrics; and Sarah Heilshorn, PhD, professor of materials science and engineering. Focusing mainly on the gastrointestinal tract, this multidisciplinary team provided proof of principle that organoids could model infectious diseases.

“With an organoid system, you can start at the infection and look at the very earliest events that occur after infection. And those can give insights as to what needs to be blocked therapeutically,” Kuo explains.

Distal Lung Organoids

After the initial success with gastrointestinal organoids, Ameen Salahudeen, MD, PhD, a hematology and oncology postdoctoral fellow working in Kuo’s lab, led efforts to expand this work by developing distal lung organoids. He partnered with lung stem cell expert Tushar Desai, MD, associate professor of pulmonary, allergy, and critical care medicine at Stanford.

The distal lung is composed of terminal bronchioles and alveolar air sacs, where inhaled air passes through the tiny ducts from the bronchioles into the elastic air sacs. It performs essential respiratory functions that can be compromised by inflammatory or infectious disorders, such as COVID-19 pneumonia.

“Growing distal lung cultures in a pure way that doesn’t require any supporting feeder cells and is in a chemically defined media had not been possible,” Kuo says. “We were able to do this very beautifully—to grow alveoli at the terminal bronchioles as long-term human cultures.”

The team developed two types of distal lung organoids. Both were made from human distal lung samples provided by Stanford cardiothoracic surgeon Joseph Schrager, MD.

They grew the first type, alveolar organoids, from single alveolar type 2 (AT2) stem cells. AT2 cells have several important functions that together help control the immune response to decrease lung injury and repair. The scientists then induced the AT2 cells to produce alveolar type 1 (AT1) cells, which are the thin-walled cells lining the alveolar air sacs; they are essential for the lung’s gas-exchange function.

“The second type are the basal organoids, which grow from single basal stem cells. They give rise to the mucus-secreting club cells and the ciliated cells with beating hair. And we can see the beating hair under the microscope—it’s quite dramatic,” describes Kuo. “That’s a very nice reproduction of the differentiation and function of the lung.” The team also grows a mixture of alveolar and basal organoids.

They selected these organoid types to determine which cell types in the bronchioles and alveoli were infectible—in hopes of identifying the different mechanisms for how viruses cause respiratory compromise.

Initially, they tested the distal lung organoids using the H1N1 influenza virus, collaborating with Stanford molecular virology expert Jeff Glenn, MD, PhD.

The team fluorescently labeled the virus and infected the lung organoids, demonstrating that the virus replicated in both basal and alveolar organoids. Next, they did more sophisticated PCR-based testing to show that the virus replicated its genome.

COVID-19 Model

“But then the COVID-19 pandemic hit, so we initiated a fabulous collaboration with infectious disease expert Catherine Blish, MD, PhD, in the Department of Medicine, to infect our lung organoids with SARS-CoV-2. This was driven by a talented MD-PhD student in my lab, Shannon Choi,” says Kuo. “She worked with Arjun Rustagi, an infectious disease fellow in Catherine Blish’s lab, who infected the organoids in a biosafety-level-3 lab.”

Another partnership was critical, though. An important coronavirus receptor, called angiotensin-converting enzyme 2, or ACE2, resides inside the lung organoids. But ACE2 needed to be on the outside of the organoid to get the infection going.

Luckily, Amieva previously devised a way to flip intestinal organoids inside out. Working together, Choi and Amieva turned the lung organoids inside out.

As reported in Nature in November 2020, the team demonstrated that the coronavirus infected their distal lung organoids, including the alveolar air sacs, where COVID-19 pneumonia originates. They also identified a new airway subpopulation as a COVID-19 virus target cell.

“Everyone knew basal cells were stem cells in the lung, but they thought they were all equivalent. Using our organoids, we discovered an unknown basal cell subpopulation containing the stem cell activity. And then we showed this subpopulation actually existed in human lungs in very interesting anatomic locations,” Kuo says.

COVID-19 Applications

According to Kuo, their distal lung organoids have three major applications for COVID-19.

They are using them to screen potential coronavirus therapeutic antibodies and to understand how these treatments work. Although initially focused on COVID-19, this screening will likely expand to other kinds of lung infections in the future.

Because the distal lung with the alveoli is the site of the COVID-19 pneumonia, they also plan to use the organoids to identify the underlying biological mechanisms behind coronavirus infection. Finally, they plan to extend their organoid system to incorporate immune cells and understand more complex processes. In particular, they plan to model the dreaded cytokine storm.

Overall, Kuo emphasizes that this organoid research represents a huge team effort involving many investigators with wide-ranging expertise from various departments at Stanford, as well as an “interesting evolution of events.” “Now we have a human experimental system to model SARS-CoV-2 infection of the distal lung with alveoli, which is the site of the lung disease that kills patients,” he summarizes. “We know patients die because of severe pneumonia and lung failure. We can now recapitulate this in the dish. So, we can study how it works, and also test drug treatments.” 

This is a reposting of my feature article in the recent Stanford Medicine Annual Report. Check it out to see videos of these lung organoids.

Innovative Antibody Treatment Proves Safe and Effective for Immune Disorders

Many blood and immune disorders could be cured by transplanting healthy blood stem cells from a matched donor. But first the patients need a pretreatment procedure to eliminate their own blood stem cells, making room in the bone marrow for the donor cells to take their place.

The problem is that the standard pretreatments—chemotherapy or radiation—are very toxic. Doctors don’t want to give them to vulnerable children, such as those with a rare genetic disorder called severe combined immunodeficiency (SCID).

Infants with SCID have compromised immune systems that struggle to fight off even common infections caused by viruses and fungi. These babies have many chronic and life-threatening problems, including frequent lung infections, chronic diarrhea, and recurrent sinus infections.

Judy Shizuru, MD, PhD, reviews data with Wendy Pang, PhD

“Without treatment, SCID infants usually die from infections within the first two years of life. Blood stem cell transplants are the only definitive cure for this disease,” says Judith Shizuru, MD, PhD, professor of blood and marrow transplantation and cellular therapy and of pediatrics. “But transplants usually involve chemotherapy, and we don’t want to give these agents to these children because they’re particularly susceptible to the damaging short-term and long-term effects—including growth defects, neurological problems, and increased risk of cancers. This is especially true for certain subtypes of SCID.”

Instead, SCID patients are often given a blood stem cell transplant without pretreating with chemotherapy to create space in their bone marrow. But then the donors’ self-renewing blood stem cells may not fully engraft, so the kids can’t robustly regenerate their immune systems. These children have to rely on regular intravenous immunoglobulin infusions to boost their immune response, and the effectiveness of donor immune cells can wane over time.

The great need for a less toxic pretreatment for blood stem cell transplants inspired Shizuru to initiate a Stanford study testing a novel antibody pretreatment in SCID patients—in collaboration with Rajni Agarwal-Hashmi, MD, associate professor of pediatrics, and other stem cell transplantation and regenerative medicine specialists at Stanford and UC-San Francisco.  

Targeting Blood Stem Cells

The novel pretreatment uses the JSP191 antibody to target a protein called CD117, found on the surface of blood stem cells. The antibody binds to this protein, which then blocks CD117 from binding to a stem cell factor critical for keeping blood stem cells alive. When the interaction between CD117 and the essential stem cell factor is interrupted, the patient’s blood stem cells are depleted—making space for the donor’s healthy cells to engraft.

“It’s not like chemotherapy or radiation,” says Shizuru. “It’s a targeted way to deplete the blood stem cells without damaging normal healthy cells.”

The Stanford team chose SCID patients for their first human JSP191 clinical trial in part because these children have a unique biology—they lack lymphocytes, so they are less likely to immunologically reject the blood stem cells from a donor. Since immune suppressive medications aren’t necessary, the researchers can more easily see if the antibody therapy clears space in the bone marrow and the transplant works.

Initially, the clinical trial studied older children and adults with SCID whose first blood stem cell transplant had failed, so that they could evaluate whether JSP191 therapy was safe and well tolerated. The participants ranged in age from 3 years old to mid-30s, but most were between 11 and 13 years old. According to Shizuru, many of these kids had chronic infections and also wanted to be liberated from having intravenous immunoglobulin infusions.  

Rajni Agarwal-Hashmi, MD

Promising Results

The results are very promising, as Shizuru reported in 2019 at the American Society of Hematology annual conference. The antibody safely created room in the patients’ bone marrow, allowing healthy donor stem cell engraftment without common side effects like transfusion reactions, treatment-related toxicities, or bone marrow suppression.

“The wonderful thing about the antibody JSP191 is it’s super-safe. This conditioning agent doesn’t affect the DNA or any other organ, as far as we can tell,” explains Shizuru. “We give it as a onetime, really low dose. And it’s not showing any side effects. It’s an amazing drug.” 

The study’s clinicians even remarked that the re-transplant kids looked bored in the hospital because the expected complications didn’t happen, says Shizuru. “The patients’ counts didn’t drop. They didn’t have increased infections. They didn’t need blood transfusions,” she says. “So, we decided to give the antibody as an inpatient treatment and then do everything else as outpatient after 48 hours.”

The results were promising from the start. The first participant pretreated with JSP191 was a 3-year-old girl with chronic diarrhea and infections. After about a year, she no longer had diarrhea and started going to school for the first time. In fact, her family was infected with COVID-19 and she did fine, as Shizuru learned during a public discussion.

Expanding the Clinical Trial

Based on the safety and success of the first phase, the JSP191 trial expanded to include infants newly diagnosed with SCID. Two infants have received the antibody pretreatment followed by a blood stem cell transplant.

The first infant did really well, demonstrating signs that his donor cells may fully restore his immune function. The second infant’s response was more complicated; the researchers determined that she had some immune function that may have rejected the maternal stem cells. She subsequently underwent another transplant without the antibody agent, using a mix of chemotherapies.  

After their initial success, Shizuru’s team expanded the use of JSP191 to include other vulnerable populations—older adults with acute myeloid leukemia (AML) or myelodysplastic syndromes (MDS). AML is a type of leukemia in which DNA mutations cause the rapid growth of abnormal cells that build up in the bone marrow. Although it starts in the bone marrow, AML often quickly moves to the blood and sometimes spreads to other parts of the body. MDS are a group of diverse bone marrow disorders in which the bone marrow does not produce enough healthy blood cells. Both AML and MDS primarily occur in people over 65 years old.

This adult study is based on the preclinical work of Wendy Pang, MD, PhD, who was a postdoctoral fellow in the Shizuru laboratory. She showed that the disease-causing MDS and AML stem cells express CD117, so they can be targeted by JSP191. Further, the team observed synergistic eradication of stem cells when these anti-CD117 antibodies were combined with low-dose radiation.

The ongoing clinical trial utilizing JSP191 combined with low-dose radiation is led by Lori Muffly, MD, assistant professor of blood and marrow transplantation and cellular therapies. The preliminary results are encouraging based on the first six participants, who were older adults (64–74 years old) with AML or MDS. The researchers observed no side effects associated with JSP191, and the patients’ blood stem cell transplants were successful.

“We transplanted our first SCID babies and then opened the trial up to older patients with AML and MDS. So, now we’re covering the full spectrum for this targeted therapy: from a 3-month-old infant with SCID to a 74-year-old with AML,” Shizuru says.

The JSP191 project has now moved to a biotechnology company, Jasper Therapeutics. Shizuru expects that in the future, the studies will expand to include sickle cell disease, a group of inherited red blood cell disorders, where the JSP191 antibody can help to engraft the donor cells.

“In terms of pretreatment, there’s been no innovation on transplant agents in decades. People have been innovating on transplant by simply reducing the dose of chemotherapies, but we haven’t seen a successful new agent,” explains Shizuru. “The development of JSP191 leverages our understanding of the biology of blood stem cells by targeting a critically important molecule. JSP191 antibody is now the platform agent.”

This is a reposting of my feature article in the recent Stanford Medicine Annual Report.

Scientists uncover surprising behavior of a fatty acid enzyme with potential biofuel applications

Derived from microscopic algae, the rare, light-driven enzyme converts fatty acids into starting ingredients for solvents and fuels.

A study using SLAC’s LCLS X-ray laser captured how light drives a series of complex structural changes in an enzyme called FAP, which catalyzes the transformation of fatty acids into starting ingredients for solvents and fuels. This drawing captures the starting state of the catalytic reaction. The dark green background represents the protein’s molecular structure. The enzyme’s light-sensing part, called the FAD cofactor, is shown at center right with its three rings absorbing a photon coming from bottom left. A fatty acid at upper left awaits transformation. The amino acid shown at middle left plays an important role in the catalytic cycle, and the red dot near the center is a water molecule. (Damien Sorigué/Université Aix-Marseille)

By Jennifer Huber

Although many organisms capture and respond to sunlight, it’s rare to find enzymes – proteins that promote chemical reactions in living things – that are driven by light. Scientists have identified only three so far. The newest one, discovered in 2017, is called fatty acid photodecarboxylase (FAP). Derived from microscopic algae, FAP uses blue light to convert fatty acids into hydrocarbons that are similar to those found in crude oil.

“A growing number of researchers envision using FAPs for green chemistry applications because they can efficiently produce important components of solvents and fuels, including gasoline and jet fuels.” says Martin Weik, the leader of a research group at the Institut de Biologie Structurale at the Université Grenoble Alpes.

Weik is one of the primary investigators in a new study that has captured the complex sequence of structural changes, or photocycle, that FAP undergoes in response to light, which drives this fatty acid transformation. Researchers had proposed a possible FAP photocycle, but the fundamental mechanism was not understood, partly because the process is so fast that it’s very difficult to measure. Specifically, scientists didn’t know how long it took FAP to split a fatty acid and release a hydrocarbon molecule.

Experiments at the Linac Coherent Light Source (LCLS) at the Department of Energy’s SLAC National Accelerator Laboratory helped answer many of these outstanding questions. The researchers describe their results in Science.

All the tools in a toolbox

To understand a light-sensitive enzyme like FAP, scientists use many different techniques to study processes that take place over a broad range of time scales. For instance, photon absorption happens in femtoseconds, or millionths of a billionth of a second, while biological responses on the molecular level often happen in thousandths of a second.

“Our international, interdisciplinary consortium, led by Frédéric Beisson at the Université Aix-Marseille, used a wealth of techniques, including spectroscopy, crystallography and computational approaches,” Weik says. “It’s the sum of these different results that enabled us to get a first glimpse of how this unique enzyme works as a function of time and in space.”

The consortium first studied the complex steps of the catalytic process at their home labs using optical spectroscopy methods, which investigate the electronic and geometric structure of atoms in the samples, including chemical bonding and charge. Spectroscopic experiments identified the intermediate states of the enzyme that accompanied each step, measured their lifetimes and provided information on their chemical nature. These results revealed the need for the ultrafast capabilities of the LCLS X-ray free-electron laser (XFEL), which can track the molecular motion with atomic precision.

A structural view of changes in the FAP molecule during the catalytic process was provided by serial femtosecond crystallography (SFX) at the LCLS. During these experiments, a jet of tiny FAP microcrystals was hit with optical laser pulses to kick off the catalytic reaction. This ensured that all the molecules react at a similar time, synchronizing their behavior and making it possible to track the process in detail. Extremely brief, ultrabright X-ray pulses then measured the resulting changes in the enzyme’s structure.

By integrating thousands of these measurements – acquired using various time delays between the optical and X-ray pulses – the researchers were able to follow structural changes in the enzyme. They also determined the structure of the enzyme’s resting state by probing without the optical laser.

Surprisingly, the researchers found that in the resting state, the light-sensing part of the enzyme has a bent shape. “This small molecule, called the FAD cofactor, is a derivative of vitamin B2 that acts like an antenna to capture photons,” Weik says. “It absorbs blue light and initiates the catalytic process. We thought the starting point of the FAD cofactor was planar, so this bent configuration was unexpected.”

The bent shape of the FAD cofactor was first discovered by X-ray crystallography at the European Synchrotron Radiation Facility, but the scientists had suspected this bend was an artifact of radiation damage, a common problem for crystallographic data collected at synchrotron light sources.

“Only SFX experiments could confirm this unusual configuration because of their unique ability to capture structural information before damaging the sample,” Weik says. “These experiments were complemented by computations. Without the high-level quantum calculations performed by Tatiana Domratcheva of Moscow State University, we wouldn’t have understood our experimental results.”

Next steps

Even with this improved understanding of FAP’s photocycle, unanswered questions remain. For example, researchers know carbon dioxide is formed during a certain step of the catalytic process at a specific time and location, but they don’t know if it is transformed into another molecule before leaving the enzyme.

“In future XFEL work, we want to identify the nature of the products and to take pictures of the process with a much smaller step size so as to resolve the process in much finer detail,” says Weik. “This is important for fundamental research, but it can also help scientists modify the enzyme to do a task for a specific application.”

Such precision experiments will be fully enabled by upcoming upgrades to the LCLS facility that will increase its pulse repetition rate from 120 pulses per second to 1 million pulses per second, transforming scientists’ ability to track complex processes like this.

Other researchers are already working towards industrial FAP applications, including a group that is designing an economic way to produce gases such as propane and butane.

The interdisciplinary consortium included researchers from the Institute of Structural Biology in Grenoble, Max Planck Institute for Medical Research in Heidelberg, Université Aix-Marseille, Ecole Polytechnique in Paris-Palaiseau, the Integrative Biology of the Cell Institute in Paris-Saclay, Moscow State University, the ESRF and SOLEIL synchrotrons in Grenoble and Paris-Saclay, and the team at SLAC National Accelerator Laboratory.

LCLS is a DOE Office of Science user facility. Major funding for this work came from the French National Research Agency (ANR).

Citation: D. Sorigué et al., Science, 9 April 2021 ((https://doi.org/10.1126/science.abd5687)

For questions or comments, contact the SLAC Office of Communications at communications@slac.stanford.edu.

This reposting of my news release, courtesy of SLAC National Accelerator Center.

Reassessing the Global Dataset of Wave Climate Projections

A snapshot of significant wave height (defined as the average of the highest 33% of waves) from a simulated realization of the future, submitted to COWCLIP 2.0. Energetic waves (yellow) originating from tropical cyclones can be seen in the North Pacific. Large waves in the southern hemisphere are due to extra-tropical storms, of large spatial extent, that continuously blast the Southern Ocean. Credit: Ben Timmermans

Wind-generated ocean waves — “wind waves” — can be major disruptors of coastal communities, marine ecosystems, offshore industries, and shipping, causing considerable environmental, geophysical, and socioeconomic impacts across the globe. Large waves during past winter storms, for example, stripped volumes of sand from Monterey, California beaches, attacked vulnerable marine terraces, and ultimately caused steep cliffs near Big Sur to crash into the sea. And around the world, these kinds of extreme weather events are becoming increasingly frequent and intense.

So it is critical to understand how global and regional wave conditions may evolve under climate change. This knowledge can then be integrated into comprehensive assessments of future coastal hazards and vulnerabilities to guide climate adaptation strategies.

Waves are generated from wind stress on the ocean surface — stronger storms generate larger waves. However, factors like storm size, intensity, translation speed, and structure combine to create different wave conditions. Modeling how atmospheric wind fields can lead to different spatial patterns of surface waves is critical for forecasts on weather time scales, but predicting how climate change alters wave conditions is much more complicated. Addressing this problem requires high performance computing.

Researchers at Lawrence Berkeley National Laboratory (Berkeley Lab) are tackling this challenge by generating and analyzing ocean wave climate projections using supercomputing resources at the National Energy Research Scientific Computing Center (NERSC), a Department of Energy user facility located at Berkeley Lab. They are also helping compile the results of international wave climate studies, creating a global ensemble dataset for widespread use by stakeholders, governments, and the research community. In the past year, two community-wide papers covering this research — including work done at Berkeley Lab — were published, one in Nature Climate Change and, more recently, in Scientific Data, also a Nature publication.

Modeling Wind-Wave Climate

Scientists use numerical general circulation models (GCMs) to simulate the dynamics and thermodynamics of the Earth’s atmosphere. Increasingly, they employ coupled GCMs to simulate the atmosphere and ocean simultaneously — and even include other components such as land hydrology — allowing feedback between the various systems. These models enable scientists to investigate the properties of the Earth’s weather and climate, both in the past and possible futures.

Only the most recent coupled atmosphere-ocean climate models include in-line wind-driven wave calculations. In addition, many climate models generate data at fairly coarse resolution, which prevents the identification of more intense storms such as tropical cyclones. The research published in Climate Change and Scientific Data compared multiple off-line wind-driven wave calculations.

Among the co-authors on these papers is Ben Timmermans, a researcher at the National Oceanography Centre in the United Kingdom and a former post-doctoral fellow in Berkeley Lab’s Climate and Ecosystems Science Division. As a postdoc, Timmermans worked with Michael Wehner, a senior scientist in Berkeley Lab’s Computational Research Division, to develop a high-resolution climate projection of average wave conditions across the globe. This work relied on simulations that Wehner had previously generated based on atmospheric data, which were collected in three-hour increments with a spatial resolution of either 25-kilometer squared or 100-kilometer squared. Simulating a 25-kilometer dataset took 64 times more computational resources than a 100-kilometer one.

“These atmospheric model calculations would have been impossible without NERSC’s computers, scratch disk, and high performance storage system,” said Wehner. “We used several hundred million hours and about 7,000 processors on Hopper and Cori for the project.”

However, Wehner’s original atmospheric simulations did not model how the atmosphere interacts with wind waves. So Timmermans extended this work to also model and analyze global wave conditions, which represented both present and possible future wave climate. NERSC again played a critical role, supplying three million core hours that ran concurrently on 20 nodes of Edison. Using the high-resolution data and NERSC computing power, the Berkeley team was able to identify tropical storms and extreme waves that the lower-resolution data lacked.

“The abundance of resources at NERSC allowed me to push the wave model almost to its limits in terms of parallel computing capability,” Timmermans said. 

Assembling Wave Climate Projections

This Berkeley Lab project is part of a new generation of global wind-wave studies completed by several international modeling groups. These individual studies, however, use various statistical approaches, dynamical wind-wave models, and data structures, making comparisons between the analyses difficult. In addition, single studies alone cannot be used to quantify total uncertainty given the range and diversity of available wind-wave modeling methods. Without a broader research effort, it remains unclear why the standalone studies sometimes differ in their projected changes in wind-wave characteristics across the world’s oceans.

The Coordinated Ocean Wave Climate Project (COWCLIP) is trying to overcome this problem by creating a consistent multivariate dataset of global wave climate projections for widespread use. Berkeley Lab is one of ten contributing institutions to COWCLIP phase 2, as described in the Scientific Data paper; all ten contributing institutions validated their global wave projection datasets with respect to observations, an important part of the production process.

For example, Timmermans’ validation involved a comparison of his projections of wind speed and wave height distributions against observations from fixed-position oceanic data buoys maintained by the National Oceanic and Atmospheric Administration. However, the COWCLIP2 team also conducted validation on the entire ensemble of datasets, comparing against 26 years of global satellite measurements of significant wave height on a global and regional scale.

“COWCLIP is a coordinated community effort to gather and explore output from state-of-the-art simulations of ocean wave climate, identifying and quantifying the key sources of uncertainty,” Timmermans said. “This new dataset will support future broad-scale coastal hazard and vulnerability assessments and climate adaptation studies in many offshore and coastal engineering applications.”

Berkeley Lab’s high-resolution climate model output used to drive the wave model is suitable for many other types of analyses and is freely available at https://portal.nersc.gov/c20c/.

This is a reposting of my news feature, courtesy of Lawrence Berkeley National Laboratory. 

Behind the scenes with a co-director of The Pride Study

In our “Behind the Scenes” series, Stanford Medicine physicians, nurses, researchers and staff members share a glimpse of their daily lives.

For Stanford obstetrician/gynecologist Juno Obedin-Maliver, MD, MPH, there is no typical day. Part of what she loves about her job is that every day is different.

Obedin-Maliver practices the full spectrum of gynecology, including outpatient, inpatient, operative and emergency services. She also co-directs The PRIDE Study, which is a national prospective, longitudinal cohort of sexual and/or gender minority people — including but not limited to lesbian, gay, bisexual, transgender and queer people.

I was excited to speak with her about how she fits all of this into her day — both before and during the COVID-19 pandemic.

Pre-COVID morning routine

I get up between 5 a.m. and 6:15 a.m. I usually make some tea and have breakfast before getting my three-year-old son up, dressed and fed. Then, either my partner or I take him to school. Next, I head down to Stanford from San Francisco where I live.

Organizing the workweek

I see patients about 30% of the time, and the rest of the time I do research. Days that I don’t see patients are a mix of research writing and meetings — with overnight calls or surgery kind of sprinkled in here and there.

Part of my research team is at Stanford, part at the University of California San Francisco and part at our office in the Oakland City Center. So, I have meetings with folks all over the Bay, and also all over the country, because we have collaborators and stakeholders across the United States.

The PRIDE Study

The main focus of  The PRIDE Study is understanding the relationship between being a sexual and/or gender minority person and a person’s health. And we think about health broadly: physical health, mental health, social health and wellbeing. We want to understand in more detail the well-documented health disparities among sexual and gender minority people, but also their health resiliency. We’ve enrolled about 18,000 people in the study.

I’m also working to build an LGBTQ+ program at Stanford, which will include clinical care, research and education.

Juno Obedin-Maliver, MD, MPH, and Mitchell Lunn, MD, co-direct The PRIDE Study, a national prospective, longitudinal cohort of sexual and/or gender minority people.
Most productive time of the day

My most productive time is in the morning at home. I usually triage my email — deleting spam, putting actionable items on my to-do list and putting anything that requires significant time on my calendar. And if I get up at 5 am, I can get an hour of uninterrupted writing in before my son wakes up, which is awesome.

Evening ritual

I get home between 6 p.m. and 7:30 p.m., then I just hang out with my son and my partner. We give him dinner and a bath, read him books and get him to sleep. And then we have our own dinner. Sometimes we just hang out until bedtime. And sometimes, unfortunately, we get back on the computer to work.

In the evening, I like to meditate, if only for 10 minutes. I remember what I’m grateful for. And I generally read a novel before I go to bed. Right now, I’m reading a book called The Hakawati by Rabih Alameddine. It’s pretty great. I try to get to sleep by 10 or 10:30 p.m.

My day during the pandemic

I still see patients one day a week, and it’s a mix of in-person and video visits in the clinic. I also work some shifts on labor and delivery.

In terms of research, my team is still rocking and rolling, despite the challenges of COVID-19 and systemic violence around the country. I’m very luck to work with an inspiring team dedicated to equity and justice.

Professionally, it’s been a productive time, and we’ve published a number of papers. We’ve also launched a survey about the impact of COVID-19 for LGBTQ+ people, and a related survey about respiratory symptoms, and have had a few thousand responses already. The pandemic seems to be exacerbating systems of inequality, and that’s certainly true for LGBTQ+ — and even more so for LGBTQ+ people of color and those who are economically disadvantaged. As we enter Pride Month, we are also about to launch our fourth annual questionnaire on June 8, and celebrate having over 18,000 participants.

Having a 3-year-old at home and splitting his care throughout the day with my partner has been a big challenge though. Our kiddo misses his friends and school, as we all do. In many ways, we’re closer than ever, and have had a lot of opportunities to do crafts and bake — and we’re growing food on our porch (tomatoes, lettuce, peppers, chard and strawberries!).

On the other hand, trying to still fit in a full work day is a struggle; it means working before he is up and long after he goes to sleep, and unfortunately more screen time for him than ever before. That being said, we’re so lucky to be healthy, have access to food and have jobs that allow us to work at least some of the time from home while still being of service.

Photos by Steve Fisch

This is a reposting of my Scope story, courtesy of Stanford School of Medicine.

Physicians re-evaluate use of lead aprons during X-rays

When you get routine X-rays of your teeth at the dentist’s office or a chest X-ray to determine if you have pneumonia, you expect the technologist to drape your pelvis in a heavy radioprotective apron. But that may not happen the next time you get X-rays.

There is growing evidence that shielding reproductive organs has negligible benefit; and because a protective cover can move out of place, using it can result in an increased radiation dose to the patient or impaired quality of diagnostic images.

Shielding testes and ovaries during X-ray imaging has been standard practice since the 1950s due to a fear of hereditary risks — namely, that the radiation would mutate germ cells and these mutations would be passed on to future generations. This concern was prompted by the genetic effects observed in studies of irradiated fruit flies. However, such hereditary effects have not been observed in humans.

“We now understand that the radiosensitivity of ovaries and testes is extremely low. In fact, they are some of the lower radiation-sensitive organs — much lower than the colon, stomach, bone marrow and breast tissue,” said  Donald Frush, MD, a professor of pediatric radiology at Lucile Packard Children’s Hospital Stanford.

In addition, he explained, technology improvements have dramatically reduced the radiation dose that a patient receives during standard X-ray films, computerized tomography scans and other radiographic procedures. For example, a review paper finds that the radiation dose to ovaries and testes dropped by 96% from 1959 to 2012 for equivalent X-ray exams of the pelvis without shielding.

But even if the radioprotective shielding may have minimal — or no — benefit, why not use it just to be safe?

The main problem is that so-called lead aprons — which aren’t made of lead anymore — are difficult to position accurately, Frush said. Even following shielding guidelines, the position of the ovaries is so variable that they may not be completely covered.  Also,  the protective shield can obscure the target anatomy. This forces doctors to live with poor-quality diagnostic information or to repeat the X-ray scan, thus increasing the radiation dose given to the patient, he said.

Positioning radioprotective aprons is particularly troublesome for small children.

“Kids kick their legs up and the shield moves while the technologists are stepping out of the room to take the exposure and can’t see them. So the X-rays have to be retaken, which means additional dose to the kids,” Frush said.

Another issue derives from something called automatic exposure control, a technology that optimizes image quality by adjusting the X-ray machine’s radiation output based on what is in the imaging field. Overall, automatic exposure control greatly improves the quality of the X-ray images and enables a lower dose to be used.  

However, if positioning errors cause the radioprotective apron to enter the imaging field, the radiographic system increases the magnitude and length of its output, in order to penetrate the shield.

“Automatic exposure control is a great tool, but it needs to be used appropriately. It’s not recommended for small children, particularly in combination with radioprotective shielding,”  said Frush.

With these concerns in mind, many technologists, medical physicists and radiologists are now recommending to discontinue the routine practice of shielding reproductive organs during X-ray imaging. However, they support giving technologists discretion to provide shielding in certain circumstances, such as on parental request. This position is supported by several groups, including the American Association of Physicists in MedicineNational Council on Radiation Protection and Measurements and American College of Radiology.

These new guidelines are also supported by the Image Gently Alliance, a coalition of heath care organizations dedicated to promoting safe pediatric imaging, which is chaired by Frush. And they are being adopted by Stanford hospitals.

“Lucile Packard Children’s revised policy on gonadal shielding has been formalized by the department,” he said. “There is still some work to do with education, including training providers and medical students to have a dialogue with patients and caregivers. But so far, pushback by patients has been much less than expected.”

Looking beyond the issue of shielding, Frush advised parents to be open to lifesaving medical imaging for their children, while also advocating for its best use. He said:

“Ask the doctor who is referring the test: Is it the right study? Is it the right thing to do now, or can it wait? Ask the imaging facility:  Are you taking into account the age and size of my child to keep the radiation dose reasonable?”

Photo by Shutterstock / pang-oasis

This is a reposting of my Scope story, courtesy of Stanford School of Medicine.

Physicists curate list of COVID-19 projects to join

As we continue to deal with the global COVID-19 pandemic, biomedical researchers are racing to understand the virus that causes the disease, to evaluate its spread, and to develop tests, treatments and vaccines.

Physicists are volunteering to assist in these efforts, using their skills in data analytics, machine learning, simulation, software, computing, hardware development and project management. And an organization called Science Responds is helping to match them with projects that need their support.

As Savannah Thais, a postdoctoral researcher in high-energy physics and a co-founder of Science Responds, reported at the April meeting of the American Physical Society, physicists are assisting with a variety of types of projects, divided into the following categories:

Epidemiology

Epidemiology is the branch of medical science that studies public health problems and events in order to understand what causes them, how they are distributed among populations and possible ways to control them. Epidemiologists investigate diverse problems including pollution, foodborne illnesses, natural disasters and infectious diseases such as COVID-19.

Science Responds is connecting physicists with epidemiological projects that are working to model how the virus that causes COVID-19 might spread. Physicists hope to help address a major problem that the experts making these models face: incorporating data from a multitude of dissimilar sources.

Thais says physicists have the background and experience needed to provide epidemiologists this kind of support.

“We don’t think physicists should be building their own epidemiological models from scratch, because they don’t have the domain expertise of an epidemiologist or biologist about infectious diseases,” she says. But “physicists can be most effective by providing their computing and statistics skills to interdisciplinary research.”

One epidemiology project Science Responds encourages volunteers to join is HealthMap, which displays data about COVID-19 cases across the globe over time via an openly accessible website and mobile app. HealthMap integrates and filters data from diverse, publicly available sources—including online news aggregators and reports from governments and agencies such as the World Health Organization—and then creates intuitive visualizations of the state of the outbreak by location.

Other modeling projects use analyses of the genomic features of previously studied viruses to help estimate unreported COVID-19 cases; integrate health and hospital resource data to inform localized risk predictions; and incorporate information from previous animal and human outbreaks to improve model accuracy.

Diagnosis

An important part of dealing with an epidemic is determining who has the disease, but shortages of testing supplies have made diagnosis a challenge. Science Responds promotes projects that are trying to address this gap in different ways.

Some projects use artificial intelligence to process visual or audio data. The project CAD4COVID, for example, builds off an existing technology that has been highly successful in diagnosing tuberculosis through the analysis of chest X-rays. The project COVID Voice Detector, on the other hand, is collecting audio recordings to develop an AI that can recognize signs of COVID-19 infection in a patient’s voice.

Other projects are building tools to predict who is likely to experience the most severe effects of COVID-19. These machine-learning-based efforts identify indicators such as markers that appear in blood tests or specific features from lung biopsies to predict the likelihood of long-term hospitalization or death.

Treatments and cures

The race to develop COVID-19 vaccines and treatments begins with understanding the physical structure of the virus. On this front, Science Responds collaborators are providing key support for an effort called Folding@Home, which uses computer simulations to map out the proteins the SARS-CoV-2 virus uses to reproduce and suppress a patient’s immune system. Physicists are helping to develop the protein-folding simulations, but they are also playing a pivotal role in looking for help from anyone with a computer that Folding@Home can use remotely to run folding simulations.

In addition, physicists are helping process the massive amount of data related to the SARS-CoV-2 genome. They’re hoping to identify molecules that are important to the growth and spread of the virus and to understand its mutations.

Science Responds collaborators are also aiding efforts to use machine learning to identify drugs that could be repurposed to treat COVID-19. For example, they are using natural-language-processing algorithms to comb through a massive database of scholarly articles, called CORD-19, for relevant ideas. Other projects are using deep-learning-based models with existing data to predict how commercially available drugs will interact with the virus.

Supporting hospitals and healthcare systems

Science Responds participants are volunteering on projects to support frontline workers who are providing medical care to COVID-19 patients. These efforts include developing models to help predict hospital overload and to allow for the sharing of resources such as mechanical ventilators and personal protective equipment.

One example is the CHIME project, which gathers information on hospital resources and predicts when the needs of patients will exceed an institution’s capacity. CHIME has already been deployed in several hospitals, including the University of Pennsylvania Health System.

Another project in this area is COVID Care Map, which is using open-source data to map existing supplies of hospital beds, ventilators and other resources needed to care for COVID-19 patients such as available staff.

Other projects highlighted by Science Responds are aimed at improving telehealth. Enhanced at-home care could reduce the spread of COVID-19 by eliminating unnecessary hospital visits and improving access to care for rural areas.

Researchers are helping to develop AI-based chatbots that can be used to assess possible infections, educate patients and call on human providers when necessary. Other projects are working to combine in-home sensors and cameras with AI-assisted technologies to remotely monitor the health of vulnerable populations.

Socio-economic response

Finally, Science Responds volunteers are also working to address what they call “second-order effects,” not directly related to healthcare.

Some projects deal with infodemiology, research into what we can learn from user-contributed, health-related content on the internet. Researchers are analyzing millions of real-time tweets related to COVID-19 to answer questions like: How are people reacting to the outbreak? How is Twitter being used to pass on vital information? How is Twitter being abused to spread false information, panic and hate?

Physicists with data-analysis and data-engineering expertise can volunteer for projects aimed at bringing attention to at-risk populations. Thais leads a project that is developing a COVID-19 Vulnerability Index, an AI-based predictive model used to identify communities at high risk of socio-economic and health impacts associated with the spread of COVID-19.

The index looks at a wide range of measures, such as whether community members have access to home Wi-Fi, whether they are affected by non-COVID health issues such as diabetes, and whether healthcare resources are available to them.

Are you a physicist looking to volunteer? Thais recommends checking out the Science Responds website, which lists projects organized by their required skills, highlights available data sources, computing resources and funding opportunities, and provides instructions for getting connected.

Illustration by Sandbox Studio, Chicago with Ana Kova

This is a reposting of my news feature, courtesy of Symmetry magazine.

 

Nerve interface provides intuitive and precise control of prosthetic hand

Current state-of-the-art designs for a multifunctional prosthetic hand are restricted in functionality by the signals used to control it. A promising source for prosthetic motor control is the peripheral nerves that run from the spinal column down the arm, since they still function after an upper limb amputation. But building a direct interface to the peripheral nervous system is challenging, because these nerves and their electrical signals are incredibly small. Current interface techniques are hindered by signal amplitude and stability issues, so they provide amputees with only a limited number of independent movements. 

Now, researchers from the University of Michigan have developed a novel regenerative peripheral nerve interface (RPNI) that relies on tiny muscle grafts to amplify the peripheral nerve signals, which are then translated into motor control signals for the prosthesis using standard machine learning algorithms. The research team has demonstrated real-time, intuitive, finger-level control of a robotic hand for amputees, as reported in a recent issue of Science Translational Medicine.

“We take a small graft from one of the patient’s quadricep muscles, or from the amputated limb if they are doing the amputation right then, and wrap just the right amount of muscle around the nerve. The nerve then regrows into the muscle to form new neuromuscular junctions,” says Cindy Chestek, an associate professor of biomedical engineering at the University of Michigan and a senior author on the study. “This creates multiple innervated muscle fibers that are controlled by the small nerve and that all fire at the same time to create a much larger electrical signal—10 or 100 times bigger than you would record from inside or around a nerve. And we do this for several of the nerves in the arm.”

This surgical technique was initially developed by co-researcher Paul Cederna, a plastic surgeon at the University of Michigan, to treat phantom limb pain caused by neuromas. A neuroma is a painful growth of nerve cells that forms at the site of the amputation injury. Over 200 patients have undergone the surgery to treat neuroma pain.

“The impetus for these surgeries was to give nerve fibers a target, or a muscle, to latch on to so neuromas didn’t develop,” says Gregory Clark, an associate professor in biomedical engineering from the University of Utah who was not involved in the study. “Paul Cederna was insightful enough to realize these reinnervated mini-muscles also provided a wonderful opportunity to serve as signal sources for dexterous, intuitive control. That means there’s a ready population that could benefit from this approach.”

The Michigan team validated their technique with studies involving four participants with upper extremity amputations who had previously undergone RPNI surgery to treat neuroma pain. Each participant had a total of 3 to 9 muscle grafts implanted on nerves. Initially, the researchers measured the signals from these RPNIs using fine-wire, nickel-alloy electrodes, which were inserted through the skin into the grafts using ultrasound guidance. They measured high-amplitude electromyography signals, representing the electrical activity of the mini-muscles, when the participants imagined they were moving the fingers of their phantom hand. The ultrasound images showed the participants’ thoughts caused the associated specific mini-muscles to contract. These proof-of-concept measurements, however, were limited by the discomfort and movement of the percutaneous electrodes that pierced the skin.

Next, the team surgically implanted permanent electrodes into the RPNIs of two of the participants. They used a type of electrode commonly used for battery-powered diaphragm pacing systems, which electrically stimulate the diaphragm muscles and nerves of patients with chronic respiratory insufficiency to help regulate their breathing. These implanted electrodes allowed the researchers to measure even larger electrical signals—week after week from the same participant—by just plugging into the connector. After taking 5 to 15 minutes of calibration data, the electrical signals were translated into movement intent using machine learning algorithms and then passed on to a prosthetic hand. Both subjects were able to intuitively complete tasks like stacking physical blocks without any training—it worked on the first try just by thinking about it, says Chestek. Another key result is that the algorithm kept working even 300 days later.

“The ability to use the determined relationship between electrical activity and intended movement for a very long period of time has important practical consequences for the user of a prosthesis, because the last thing they want is to rely on a hand that is not reliable,” Clark says.

Although this clinical trial is ongoing, the Michigan team is now investigating how to replace the connector and computer card with an implantable device that communicates wirelessly, so patients can walk around in the real world. The researchers are also working to incorporate sensory feedback through the regenerative peripheral nerve interface. Their ultimate goal is for patients to feel like their prosthetic hand is alive, taking over the space in the brain where their natural hand used to be.

“People are excited because this is a novel approach that will provide high quality, intuitive, and very specific signals that can be used in a very straightforward, natural way to provide high degrees of dexterous control that are also very stable and last a long time,” Clark says.

Read the article in Science Translational Medicine.

Illustration of multiple regenerative peripheral nerve interfaces (RPNIs) created for each available nerve of an amputee. Fine-wire electrodes were embedded into his RPNI muscles during the readout session. Credit: Philip Vu/University of Michigan; Science Translational Medicine doi: 10.1126/scitranslmed.aay2857

This is a reposting of my news brief, courtesy of Materials Research Society.

Why do viruses like the coronavirus sometimes steal our sense of smell?

When you catch a severe cold, your nose stuffs up, you can’t smell anything and food tastes funny. Fortunately, most people regain their sense of smell once the cold runs its course. But for others, the complete (anosmia) or partial (hyposmia) loss of the sense of smell is permanent.

I spoke with Zara Patel, MD, a Stanford associate professor of otolaryngology, head and neck surgery, and director of endoscopic skull base surgery, to learn more about her research on olfactory disorders. In particular, we discussed her recent study on the possible association between post-viral olfactory loss and other cranial neuropathies, which are disorders that impair your nerves and ultimately your ability to feel or move. She also described how her work pertains to the COVID-19 pandemic.  

How does a virus impair someone’s sense of smell?

A variety of viruses can attack the cranial nerves related to smell or the mucosal tissue that surrounds those nerves. Cranial nerves control things in our head and neck — such as the nerves that allow us to speak by using our vocal cords, control our facial motion, hear and smell.

For example, COVID-19 is just one type of disease caused by a coronavirus. There are many other types of coronaviruses that cause colds and upper respiratory illnesses, as well as rhinoviruses and influenza viruses. Any of these viruses are known to cause inflammation, either directly around the nerve in the nasal lining or within the nerve itself. When the nerve is either surrounded by inflammatory molecules or has a lot of inflammation within the nerve cell body, it cannot function correctly — and that is what causes the loss or dysfunction of smell. And it can happen to anyone: young and old, healthy and sick.

How did your study investigate olfactory loss?

In my practice, I see patients who have smell dysfunction. But I’m also a sinus and skull base surgeon, so I have a whole host of other patients with sinus problems and skull-based tumors who don’t have an olfactory loss. So we did a case-control study to compare the incidence of cranial neuropathies — conditions in which nerves in the brain or brain stem are damaged — in two patient groups. Ninety-one patients had post-viral olfactory loss and 100 were controls; and they were matched as closely as possible for age and gender.

We also looked at family history of neurologic diseases — such as Alzheimer’s disease, Parkinson’s disease and stroke.

What did you find?

Patients with post-viral olfactory loss had six-times higher odds of having other cranial neuropathies than the control group — with an incidence rate of other cranial nerve deficits of 11% and 2%, respectively. Family history of neurologic diseases was associated with more than two-fold greater odds of having a cranial nerve deficit. Although we had a small sample size, the striking difference between the groups implies that it is worthwhile to research this with a larger population.

Our findings suggest that patients experiencing these pathologies may have inherent vulnerabilities to neural damage or decreased ability of nerve recovery — something beyond known risk factors like age, body mass index, co-morbidities and the duration of the loss before intervention. For example, there may be a genetic predisposition, but that is just an untested theory at this point.

How does this work pertain to COVID-19?

Smell loss can be one of the earliest signs of a COVID-19 infection. It can sometimes be the only sign. Or it can present after other symptoms. Although it may not affect every patient with COVID-19, loss of smell and taste is definitely associated with the disease. In some countries, including France, they’ve used this as a triage mechanism. People need to know that these symptoms can be related to the COVID-19 disease process so they aren’t going about their lives like normal and spreading the virus.

The pandemic also might impact how we treat patients with olfactory dysfunction in general. When someone has a viral-induced inflammation of the nerve, we sometimes treat it with steroids to decrease the inflammation. But treating COVID-19 patients with steroids might be a bad idea because of its effect on the inflammatory processes going on in their heart and lungs.

What advice do you have for people who have an impaired sense of smell?  

First, if you lose your sense of smell and it isn’t coming back after all the other symptoms have gone away, seek care as soon as possible. If you wait too long, there is much less that we can do to help you. Interventions, including olfactory training and medications, are more effective when you are treated early.

Second, if you lose your sense of smell or taste during this pandemic and you don’t have any other symptoms, contact your doctor. The doctor can decide whether you need to be tested for COVID-19 or whether you need to self-isolate to avoid being a vector of the virus in your family or community.

Image by carles

This is a reposting of my Scope story, courtesy of Stanford School of Medicine.

Harnessing AMReX for Wind Turbine Simulations

ECP ExaWind Project Taps Bereley Lab’s AMReX to Help Model Next-Generation Wind Farms

Driving along Highway 580 over the Altamont Pass in Northern California, you can’t help but marvel at the 4,000+ wind turbines slowly spinning on the summer-golden hillsides. Home to one of the earliest wind farms in the United States, Altamont Pass today remains one of the largest concentrations of wind turbines in the world. It is also a symbol of the future of clean energy.

Before utility grids can achieve wide-scale deployment of wind energy, however, they need more efficient wind plants. This requires advancing our fundamental understanding of the flow physics governing wind-plant performance.

ExaWind, a U.S. Department of Energy (DOE) Exascale Computing Project, is tackling this challenge by developing new simulation capabilities to more accurately predict the complex flow physics of wind farms. The project entails a collaboration between the National Renewable Energy Laboratory (NREL), Sandia National Laboratories, Oak Ridge National Laboratory, the University of Texas at Austin, Parallel Geometric Algorithms, and — as of a few months ago — Lawrence Berkeley National Laboratory (Berkeley Lab).

“Our ExaWind challenge problem is to simulate the air flow of nine wind turbines arranged as a three-by-three array inside a space five kilometers by five kilometers on the ground and a kilometer high,” said Shreyas Ananthan, a research software engineer at NREL and lead technical expert on the project. “And we need to run about a hundred seconds of real-time simulation.” 

By developing this virtual test bed, the researchers hope to revolutionize the design, operational control, and siting of wind plants, plus facilitate reliable grid integration. And this requires a combination of advanced supercomputers and unique simulation codes.

Unstructured + Structured Calculations

The principle behind a wind turbine is simple: energy in the wind turns the turbine blades, which causes an internal gearbox to rotate and spin a generator that produces electricity. But simulating this is complicated. The flexible turbine blades rotate, bend, and twist as the wind shifts direction and speed. The yaw and pitch of these blades are controlled in real time to extract as much energy as possible from a wind event. The air flow also entails complex dynamics  — such as influences from the ground terrain, formation of a turbulent wakefield downstream from the blades, and turbine-turbine interactions.

To improve on current simulations, scientists need more computing power and higher resolution models that better capture the crucial dynamics. The ExaWind team is developing a predictive, physics-based, and high-resolution computational model — progressively building from petascale simulations of a single turbine toward exascale simulations of a nine-turbine array in complex terrain.

A Nalu-Wind solution to the differential equations of motion for a wind turbine operating in uniform air flow (moving from left to right). Two of the three wind turbine’s blades are pictures (think blue rectangles on left). The slice in the background represents the contours of the whirling air’s motion, showing the vertical direction of the wake structure behind the turbine blades (red indicates swirl in counterclockwise direction and blue clockwise direction around blade tip).

“We want to know things like the air velocity and air temperature across a big three-dimensional space,” said Ann Almgren, who leads the Center for Computational Sciences and Engineering in Berkeley Lab’s Computational Research Division. “But we care most about what’s happening right at the turbines where things are changing quickly. We want to focus our resources near these turbines, without neglecting what’s going on in the larger space.”

To achieve the desired accuracy, the researchers are solving fluid dynamics equations near the turbines using a computational code called Nalu-Wind, a fully unstructured code that gives users the flexibility to more accurately describe the complex geometries near the turbines, Ananthan explained.

But this flexibility comes at a price. Unstructured mesh calculations have to store information not just about the location of all the mesh points but also about which points are connected to which. Structured meshes, meanwhile, are “logically rectangular,” which makes a lot of operations much simpler and faster.

“Originally, ExaWind planned to use Nalu-Wind everywhere, but coupling Nalu-Wind with a structured grid code may offer a much faster time-to-solution,” Almgren said.

Enter AMReX

Luckily, Ananthan knew about Berkeley Lab’s AMReX, a C++ software framework that supports block-structured adaptive-mesh algorithms for solving systems of partial differential equations. AMReX supports simulations on a structured mesh hierarchy; at each level the mesh is made up of regular boxes, but the different levels have different spatial resolution.

Ananthan explained they actually want the best of both worlds: unstructured mesh near the turbines and structured mesh elsewhere in the domain. The unstructured mesh and structured mesh have to communicate with each other, so the ExaWind team validated an overset mesh approach with an unstructured mesh near the turbines and a background structured mesh. That’s when they reached out to Almgren to collaborate.

“AMReX allows you to zoom in to get fine resolution in the regions you care about but have coarse resolution everywhere else,” Almgren said. The plan is for ExaWind to use an AMReX-based code (AMR-Wind) to resolve the entire domain except right around the turbines, where the researchers will use Nalu-Wind. AMR-Wind will generate finer and finer cells as they get closer to the turbines, basically matching the Nalu-Wind resolution where the codes meet. Nalu-Wind and AMR-Wind will talk to each other using a coupling code called TIOGA.

Even with this strategy, the team needs high performance computing. Ananthan’s initial performance studies were conducted on up to 1,024 Cori Haswell nodes at Berkeley Lab’s National Energy Research Scientific Computing Center (NERSC) and 49,152 Mira nodes at the Argonne Leadership Computing Facility.

“For the last three years, we’ve been using NERSC’s Cori heavily, as well as NREL’s Peregrine and Eagle,” said Ananthan. Moving forward, they’ll also be using the Summit system at the Oak Ridge Leadership Computing Facility and, ultimately, the Aurora and Frontier exascale supercomputers — all of which feature different types of GPUs: NVIDIA on Summit (and NERSC’s next-generation Perlmutter system), Intel on Aurora, and AMD on Frontier. 

Although Berkeley Lab just started partnering with the ExaWind team this past fall, the collaboration has already made a lot of progress. “Right now we’re still doing proof-of-concept testing for coupling the AMR-Wind and Nalu-Wind codes, but we expect to have the coupled software running on the full domain by the end of FY20,” said Almgren.

NERSC is a DOE Office of Science user facility.

Top figure: Some of the 4000+ wind turbines in Northern California’s Altamont Pass wind farm. Credit: David Laporte

This is a reposting of my news feature, courtesy of Berkeley Lab.