Powered by AI
By Jennifer Racz
Artificial intelligence has burst into the public imagination through chatbots, image generators, and content creation, but in the labs of the Weizmann Institute, it is being used for something far more profound: solving some of the most pressing challenges in human and planetary health. Supported by a strong knowledge base, access to cutting-edge tools, and dedicated funding, Weizmann researchers are leveraging the incredible power of AI into a force for advancing sustainability science.
Breath of knowledge
If you live in a major city like New York, Seoul, or Tel Aviv, you can check the air quality—the concentration of particulates, levels of ozone, and whether it’s safe to breathe the air—in a matter of seconds on your phone. But for billions of people worldwide, no such data or resources exist. Monitoring air pollution requires the kind of ground tracking that is lacking in many parts of the world.
In Prof. Yinon Rudich’s group in the Department of Earth and Planetary Sciences, PhD student Nati Ofir is using his extensive training in computer science at the Weizmann Institute (where he completed both MSc and PhD degrees), to bridge this gap. Merging his background in computer science with chemistry and sustainability concepts, he uses satellite and meteorological data on atmospheric conditions above the Korean peninsula and North America, alongside information on air quality from ground stations in cities, to design models that can predict air quality where ground data is missing.
“The satellites provide images of the atmosphere, but not what’s happening on the surface,” Dr. Ofir explains. “Combining satellite images and ground data from other cities, AI learns how to make that leap.”
Given its ability to learn patterns and fill in missing information, AI has become a powerful tool in climate science, helping researchers build faster and more accurate climate models, improving weather pattern simulations, and filling in temporal or spatial gaps in observations.
“AI lets us sift through massive, complex datasets to uncover hidden patterns in the climate system,” says Prof. Rudich. “These insights open new paths for discovery and can make our predictions about the planet’s future more powerful.”
In Dr. Ofir’s case, with a laptop and several years of archived satellite data, he developed a model that can generate real-time predictions about air pollution events and risk in surrounding areas, in places such as Vietnam or Thailand—countries with limited infrastructure but high exposure.
In the future, his model could use data from low-cost satellites to predict air quality and air pollution events in cities and towns that lack monitoring infrastructure altogether.
“Satellite images are cheap compared to ground stations and can cover many countries. Once I teach the algorithm, I don’t need the ground station anymore,” he explains.
For policymakers, these forecasts could shape interventions to reduce exposure to polluted air. For families, they could protect children on high-pollution days. “AI allows us to fill in missing data,” adds Dr. Ofir, “giving everyone access to the same critical information about the air they breathe.”
Cracking the Rubisco code
If air pollution is a modern challenge, Rubisco is an ancient puzzle. This enzyme, found in both plants and bacteria, and with origins that date back to more than three billion years ago, sits at the very heart of life: it captures carbon dioxide from the atmosphere and transforms it into sugars during photosynthesis. Every carbon atom in your body, every bite of food you’ve ever eaten, has passed through the Rubisco enzyme.
Yet despite its central role, this enzyme is astonishingly inefficient. It works slowly—just one reaction per second, while other enzymes catalyze thousands—and often makes mistakes, confusing oxygen for carbon dioxide and creating toxic byproducts. Scientists have worked to engineer a faster, more accurate Rubisco enzyme for decades, but its complexity and resistance to change have made it nearly impossible.
Dr. Dina Listov, a postdoctoral fellow in Prof. Sarel Fleishman’s lab in the Department of Biomolecular Sciences, believes that AI-based tools hold the key to cracking the Rubisco code. Taking advantage of natural diversity within the enzyme’s family—plant Rubisco is better adapted to today’s oxygen-rich atmosphere, while bacterial Rubisco often works much faster—the Fleishman group sequences and compares the Rubisco proteins to identify features that could improve the enzyme. Their goal: combine the desired fragments to create a Rubisco that outperforms nature’s own.
Artificial intelligence programs are accelerating this process. While it once took months to predict whether a new hybrid would fold into a stable, functional protein, AI tools such as AlphaFold—developed by Google DeepMind—can predict 3D structures in minutes. “What used to take me half a year, I can now do in 10 minutes,” Dr. Listov says.
Instead of painstakingly stitching together Rubisco fragments through months of computation, she can now design hundreds of thousands of novel Rubisco enzymes in under an hour. AI not only accelerates her work—it opens new avenues, suggesting pathways human intuition might miss. Success could open the door to the engineering of microbes that pull carbon straight from the air and convert it into fuel or medicine.
According to Prof. Fleishman, “AI enables us to weave together vast amounts of computational and experimental data in ways we simply couldn’t before. We can move beyond nature’s existing toolkit of proteins and start designing entirely new ones—from powerful therapeutics to enzymes that can drive sustainability solutions.”
For Dr. Listov, AI has become a partner in discovery. “It doesn’t just make research faster,” she reflects. “It makes possible what was impossible.”
Following the chemical trail
Modern agriculture depends on herbicides and pesticides to secure global food supplies. These same chemicals often linger in our food, water, and bodies, raising profound public health questions about links between environmental pollutants and disease. At the Weizmann Institute, Prof. Eran Segal, from the Departments of Molecular Cell Biology and Computer Science and Applied Mathematics, and Dr. David Zeevi, from the Department of Plant and Environmental Sciences, are leading a collaborative project to untangle how these compounds travel from farm fields into humans—using AI to help reveal both their risks and possible solutions.
By analyzing blood samples from 2,500 participants alongside detailed diet records, the scientists discovered that traces of a common weed killer, quinmerac, appeared more often among vegetarians and vegans. Other analyses showed clear links between dietary choices, pesticide exposure, and metabolic markers. Such findings highlight how even health-minded individuals can be exposed to hidden risks.
Beyond human blood, Prof. Segal and Dr. Zeevi are looking at the microbes that populate lakes, soils, and guts. Dr. Zeevi describes these microbes as Earth’s “tiny witnesses.” Trillions of species are constantly metabolizing and adapting to their surroundings, leaving behind a genetic record of the pressures they face. Among them are bacteria capable of breaking down persistent agricultural chemicals.
“Analyzing the microbes is like interviewing the witnesses,” he explains. “If we learn how to read them, they can tell us what’s happening to the planet.”
AI allows scientists to “read the witnesses,” integrating massive datasets—from human metabolomics and diet logs to microbial DNA and evolutionary signatures—to train models that reveal broad rules. In one study, the group identified DNA fragments tied to pesticide breakdown, creating rules that uncovered similar patterns in other ecosystems.
Applying these models to Canadian lakes polluted by farm runoff, they found bacteria whose genes had adapted to stress—many linked to the degradation of glyphosate, one of the world’s most common herbicides. These insights are bringing researchers closer to practical solutions, from dietary changes that reduce risk to harnessing bacteria that clean the environment.
According to Dr. Zeevi, the Weizmann Institute’s supportive environment makes such ambitious work possible. Strong core facilities in chemistry and sequencing, together with the Institute’s AI hub, provide the infrastructure; its collaborative culture brings together the needed expertise; and grants from the Institute for Environmental Sustainability provide the funding that can often be hard to secure. “At Weizmann,” he says, “all you need is the idea.”
Yinon Rudich is supported by:
The Abisch-Frenkel RNA Therapeutics Center
The Ilse Katz Institute for Material Sciences and Magnetic Resonance Research
The Nancy and Stephen Grand Research Center for Sensors and Security
The Irving and Cherna Moskowitz Center for Nano and Bio Nano Imaging
Sarel Fleishman is supported by:
The Artificial Intelligence for Smart Materials Research Fund, in Memory of Dr. Uriel Arnon
David Zeevi is supported by:
The Roden Family Research Fund for Environmental Sustainability
Eran Segal is supported by:
A private donation managed by the Goldman Sachs Foundation
The Moross Integrated Cancer Center
The Sagol Institute for Longevity Research
The Swiss Society Institute for Cancer Prevention Research
The Louis H. Sackin Research Fellow Chair in Computer Science supports a staff scientist in Prof. Segal’s Lab