A protein by-product from corn processing could help cut the energy cost of industrial chemical purification. Developed by KAUST researchers, the protein can be made into a biodegradable nanofiltration membrane that separates mixtures of industrial chemicals using a fraction of the energy of traditional purification methods^[1].

Chemical separations account for a large share of global industrial energy consumption, typically through large-scale heat-driven processes such as evaporation, drying, or distillation. Molecular sieve membranes are a promising, energy-efficient alternative for industrial chemical purification, using selective nanofiltration rather than heat-driven separation. Adopting membrane technology can cut the carbon emissions of a chemical purification step by up to 90 percent.

Yet despite this promise to decrease industrial energy consumption, nanofiltration membranes themselves are not very green, says Claudia Oviedo, a Ph.D. student in Gyorgy Szekely’s lab, who led the research.

“Most membranes are based on synthetic polymers derived from fossil resources, which raises concerns about their sustainability and long-term environmental persistence,” Oviedo explains. “The motivation of our research is to reduce the dependence on fossil-based membrane building blocks by exploring bio-based alternatives, preferably made from abundant agricultural waste.”

An ideal candidate could be a protein called zein, a widely available byproduct of the global corn processing industry, the team showed. “Zein is particularly promising for membrane materials due to its hydrophobicity,” Oviedo says. This low water solubility was the key to turning zein into a solvent-resistant nanofiltration membrane.

The researchers made their bio-based membranes by dissolving the zein in a green solvent, casting this solution into a thin film, then adding water as a nonsolvent. As water displaced the solvent and the hydrophobic protein chains were squeezed together, they experienced a phenomenon called macromolecular crowding.

“Macromolecular crowding occurs when large molecules are present in high concentrations,” Oviedo says. Repelled by the water, the zein protein chains are forced to interact and organize, packing closely together to form a well-defined structured membrane.

Collaborating with Satoshi Habuchi and his KAUST team, the researchers tracked the process in microscopic detail by adding nanoscale luminous tracers called quantum dots to the starting protein solution.

Once water was added, the quantum dots’ motion gradually separated into two bands, with fast-moving dots in the upper layer and slow-moving dots in the lower layer where the membrane was forming. “This motion indicated the formation of zein-rich and zein-poor domains as the membrane structure developed,” Oviedo says.

Computational molecular dynamics simulations showed that hydrophobic protein-protein interactions drove the zein chains to pack closely and entangle, creating a crowded network where nanoscale pores are surrounded by densely arranged protein chains.

The performance of the new zein membrane’s nanofiltration was competitive with commercial synthetic nanofiltration membranes, the team found, and they used their sustainable membrane material to separate a toxic impurity from a pharmaceutical product.

“Our trials also showed that zein membranes biodegraded rapidly, in contrast to conventional fossil-based membranes that persist in the environment post-disposal,” Szekely says.

The team’s next goal is to evaluate whether bio-based nanofiltration membranes provide lower environmental impacts from cradle to grave. “Our ongoing research focuses on life cycle assessment comparing zein extraction and membrane fabrication with fossil-based membranes,” Szekely says.

 

A thin-film, flexible optical device that mimics the way the human brain senses and interprets visual information has been developed by KAUST researchers^[1]. This “optical synapse” may help address the growing need for more efficient artificial vision systems.

“Today’s cameras and computers usually separate sensing, memory, and processing into different parts, which requires data to move back and forth constantly, wasting both time and energy,” explains Manoj Kumar Rajbhar, who worked on the project supervised by Nazek El-Atab. “Our goal was to move closer to the human visual system, where sensing and processing are tightly linked.”

Previous designs for light-sensitive synaptic devices required both electrical and optical signals to operate. They were composed of complex ‘stacks’ of materials, or utilized unstable compounds such as perovskites or black phosphorus. Rajbhar and colleagues designed a synaptic device controlled entirely by light, improving energy efficiency and eliminating the need for electrical signals. Exposure to different colors of light can strengthen or weaken their device’s response, similar to how synapses in the human brain behave during learning.

“In the brain, synapses don’t only become stronger, they can also become weaker,” says Rajbhar. “That balance is essential for learning new information, filtering noise, forgetting unimportant signals, and adapting to changing environments.”

For example, the researchers imitated the classic Pavlov’s dog experiment, in which a dog learns to associate the sound of a bell with food and begins salivating in response. By using one wavelength of light to represent a bell and another to represent food, they trained the device to associate the ‘bell’ signal with ‘food’ and initiate a ‘salivation’ response.

“This type of training is especially valuable for future machine vision and neuromorphic hardware, because it allows the device to process visual information in a more brain-like way,” says Rajbhar. “Practically, this helps systems become more adaptive and better at tasks such as recognition, memory formation, and decision-making without relying on separate sensing and processing units.”

The team also simplified their device’s structure by using an ultrathin layer of manganese oxide on a flexible silicon substrate. The device works even when the silicon substrate is bent. Manganese oxide is relatively abundant and affordable, supporting future scalability. It is also more environmentally friendly than some other potential materials and hosts multiple oxidation states.

“These mixed oxidation states help create controllable defect states and oxygen vacancies, which are important for tuning how the device stores and modulates information,” says Rajbhar. “We consider these features to play an important role in the light response and memory behavior of the device.”

Their device is capable of real-time image detection and processing, and preforms logical operations compatible with existing semiconductor computing.

“This device could be useful where there is a need for lightweight, low-power systems that can sense visual information and process it locally,” says El-Atab. “Our results support advances in AI hardware, robotics, wearable electronics, and artificial vision. The study also aligns with Saudi Arabia’s Vision 2030 goals in advanced technology and semiconductor research.”

 

An intelligent optoelectronic transistor can not only sense light but also remember and learn from what it has seen. Developed by researchers at KAUST and Peking University, the “optoelectronic synapse” uses an organic material called gDPP-MeOT2 — a light-absorbing polymer that can also store and transport ions^[1].

“A conventional photodetector simply turns light into an electrical signal,” says Yazhou Wang, a postdoc in Sahika Inal’s lab who led the research. “Our device goes a step further – it can process and remember what it sees. Light triggers not only electronic signals but also ionic motion in the material, allowing the device to dynamically adjust its response over time, similar to how synapses in the brain strengthen or weaken with experience,” explains Wang.

By integrating the gDPP-MeOT2 material into an electrically controllable transistor architecture, Inal’s team, in collaboration with Nazek El-Atab’s group at KAUST, showed that the device can mimic key functions of a neural synapse in the human brain. The researchers demonstrated behaviors linked to learning and memory, including the ability to strengthen responses to repeated light signals and transition from short-term to long-term memory. These functions operate at an very low programming voltage of just 0.4 volts and on the timescale of seconds.

Using these capabilities, the team carried out optical logic operations, such as switching between binary states using light, as well as image processing tasks including image denoising through adaptive filtering.

“We were inspired by how the human retina seamlessly combines sensing, processing, and memory in a single energy-efficient system. In contrast, today’s electronics typically separate these functions, which increases complexity and power consumption,” explained Wang. “Our goal is to bring these capabilities within a single material platform, enabling devices that can not only detect light but also learn from it and adapt, much like biological vision systems.”

The device responds to both visible and near-infrared light, covering wavelengths from around 455 nm to 1100 nm. This makes it suitable for vision-related technologies that require fast, built-in image processing. Potential applications include artificial retinas, wearable sensors and autonomous systems that must interpret visual information in real time. The aim is to enable efficient, brainlike sensing and computing without relying on cloud-based processing.

According to Wang, the key challenges are to further speed up the technology and scale it into large, high-density arrays for practical implementation.

“Looking ahead, we aim to make these devices even more ‘brain-like’ by introducing functionalities such as selective responses to different chemical signals,” says Inal. “We are also working to integrate many of these synapses into large arrays, to enable real-time, on-chip visual processing. Ultimately, we hope to build systems that can sense, learn, and make decisions directly at the hardware level.”

 

A gas sensor that can provide an early warning of battery failure could improve the safety of lithium-ion batteries, used in everything from electric vehicles to grid-scale storage systems.

Lithium-ion batteries are transforming the global energy landscape, but they also come with risks. If a battery is damaged or overheated, it can trigger an unstoppable “thermal runaway” reaction that releases flammable and toxic gases and leads to catastrophic battery failure.

KAUST researchers have developed a sensor designed to detect those gases at a very early stage before conditions escalate^[1]. “Our ultimate goal is to demonstrate a closed-loop early warning and response system that could serve as the basis for a practical safety architecture in battery packs,” says Aamir Farooq, who led the team.

When a battery gets too hot, the chemicals inside start to break down and release gases. As pressure builds, a safety valve ruptures to vent them.

This breakdown also generates heat, potentially causing a chain reaction that accelerates gas formation and can ultimately cause the battery to catch fire or explode.

Among the most hazardous gases released during battery failure is hydrogen fluoride (HF), a highly toxic and corrosive compound. Previous studies have shown that HF can be detected during safety venting, before the thermal runaway process takes hold. However, conventional methods for measuring HF rely on bulky equipment and tend to underestimate emissions.

The KAUST team developed a more compact gas-sensing technique called tunable diode laser absorption spectroscopy (TDLAS). Positioned directly above the safety vent, the system provides a much more accurate HF reading.

The researchers tested the TDLAS sensor on two types of lithium-ion batteries, commonly found in laptops and electric vehicles. One type uses a cathode containing nickel, manganese and cobalt (NMC), while the other has a lithium iron phosphate (LFP) cathode. These batteries, charged to either 50 percent or 100 percent of their capacity, were heated to trigger thermal runaway.

The NMC batteries reached thermal runaway at 174–215 °C, while LFP cells were more stable, doing so at 242–249 degrees Celsius. Both types released a small burst of HF during initial safety venting, followed by a much larger burst once thermal runaway had started. “These concentrations of HF are orders of magnitude above the level that is immediately dangerous to life and health,” says Ahmad Alsewailem, a Ph.D. student in the team who led the experimental work.

Crucially, the time between the early and later bursts was about one minute for NMC cells and five minutes for LFP cells. That would be long enough for a miniaturized TDLAS sensor to act as an early warning system in electric vehicle battery packs or grid-scale energy storage systems.

“A single sensor positioned within a battery pack enclosure could monitor the gas environment continuously and trigger an alarm or automatic shutdown the moment HF is detected above a threshold level,” says Janardhanraj Subburaj, a research scientist in the team. “This is faster than other thermal runaway sensors, giving a warning window before the more violent thermal runaway phase.”

The researchers now plan to study how overcharging and mechanical damage can cause similar gas releases, and hope to couple their HF sensor with an automated battery safety system.

Advances in DNA sequencing have expanded our view of the microbial world, but the inability to cultivate most microbes has been a major constraint. Now, a systematic, predictive framework that combines existing genomic and computational modeling approaches to accelerate the discovery and cultivation of novel prokaryotic taxa has been proposed by KAUST researchers, in collaboration with an international team of scientists^[1].

“Prokaryotic microbes — bacteria and archaea — are extraordinary microorganisms with unique capabilities that are already proving useful in industrial settings and biotechnological applications,” says Diego Javier Jiménez Avella, who led the perspective study within the research group of Alexandre Rosado at KAUST. “For example, cultivated microorganisms can be used in bioremediation processes, generating bioactive compounds, and enhancing plant growth in agricultural systems.”

Current DNA sequencing technologies have unveiled an enormous diversity of microbes on a global scale, but many potentially useful prokaryotic species remain undiscovered and uncultivated. Isolating and cultivating novel microbial species remains a major bottleneck in microbiology, notes Jiménez.

“We are at a very interesting time in microbiology, where we are revealing a vast microbial world through DNA, but we still cannot access much of it experimentally,” says Rosado. “Without cultivation, we cannot properly understand how these organisms function, or translate that knowledge into applications. There is also an urgency here: environments across the globe are changing rapidly, and we risk losing microbial diversity before we have had a chance to characterize it.”

Cultivation challenges

Scientists still do not fully understand the environmental and nutritional requirements needed to sustain microbial species in laboratory settings. Each microbe requires a unique combination of factors to grow, and this challenge is compounded by competition and interactions between microbial species.

“A wide range of techniques have been developed to isolate and cultivate microbial species from natural environments, but most of these rely on set nutrient and environmental conditions that tend to favor fast-growing, heterotrophic microorganisms,” says Jiménez. “Moreover, conventional isolation methods often disrupt obligate microbial interactions,” adds Jiménez, referring to the close interdependent relationships on which some species depend.

Many microbes live in diverse symbiotic communities and rely on interactions between species to survive. These communities are often disrupted by isolation techniques, causing scientists to lose the more sensitive, rare species — or miss them entirely.

“In many cases, we are simply not recreating the right ecological context,” says Rosado. “The limitation is not the methods themselves, but the fact that they were never designed to capture the full diversity we now know exists.”

Genome-based cultivation strategies have improved knowledge of the nutritional requirements needed to cultivate specific prokaryotic taxa. However, many prokaryotic genomes contain genes with unknown or poorly annotated functions, making it more difficult to understand the microbes’ physiology and predict their metabolic needs.

“Our new approach tries to reduce these challenges by using genomic information to guide cultivation, moving away from trial and error towards more informed hypotheses,” says Jiménez. “What we propose is not a new tool, but a way to connect existing approaches into a more systematic and predictive framework.

Harnessing technological advances

Analyzing and understanding thousands of microbial species in their natural environments requires a combination of multidisciplinary techniques. In their latest paper, the research team proposes combining direct environmental DNA analyses with genome-based metabolic modeling, physiological inference, and the design of tailored growth media to improve the targeted cultivation and isolation of novel taxa.

The team first proposes targeted perturbations in natural ecosystems to increase the abundance of rare or hidden microorganisms that may otherwise remain undetected. Once these microbes have been identified and collected, their genomes can be sequenced, reconstructed, and annotated. Scientists can then infer each microbe’s metabolism and potential growth requirements. Finally, this information can be used, together with AI and machine learning, to design better cultivation strategies.

However, even with a robust, reproducible framework in place, significant challenges remain, say Jiménez and Rosado.

“A large fraction of genes in microbial genomes have unknown functions, which limits how accurately we can predict physiology. While AI and machine learning are helpful, especially for gene annotation and metabolic prediction, they are not shortcuts. They depend on good data and still require experimental validation.”

Wide-reaching potential applications

This study brings together microbiology, bioinformatics, and expertise in AI and machine learning at KAUST, and represents an important step toward improving microbial cultivation.

“Once you bring new microbes into culture, you can understand their biology and begin to use it,” says Rosado. “This can translate into environmental biotechnology, agriculture, and medical applications. For example, microbes that degrade pollutants or plastics, or that support plant growth under extreme conditions, are particularly relevant in regions like Saudi Arabia.”

“We have already identified previously uncharacterized microbial taxa, discovered novel enzymes, and generated new insights into environmental microbiomes from Saudi Arabia and beyond,” says Jiménez.

“Building on these advances, the opportunity ahead is to translate this knowledge into scalable solutions,” concludes Rosado. “From human health and desert agriculture to mangrove restoration, Red Sea ecosystems, and space biology, microbiome-based approaches could support sustainable bio-based approaches aligned with KAUST priorities in energy, water, food, and health, as well as Saudi Arabia’s Vision 2030.”

A simple yet efficient method of generating models of early human embryo development offers a robust platform for investigating infertility and developmental disorders. Developed by an international team led by KAUST, the model is already providing biological insights into the earliest stages of blastocyst formation^[1].

“We were motivated by a fundamental question in biology: how does a single cell organize into a complex embryo during the earliest stages of development?” says Mo Li, who supervised the study. “These stages, particularly blastocyst formation, are critical for implantation and successful pregnancy, yet they remain poorly understood.”

Amid global fertility decline and increasing reliance on treatments such as in vitro fertilization (IVF), there is a clear need to develop reproducible, ethical models of blastoids – early embryo-like structures – from human naïve pluripotent stem cells.

“Many IVF failures occur at or before the blastocyst stage, highlighting the need to better understand early embryonic mechanisms and improve clinical outcomes,” says Arun Pandian Chandrasekaran, a postdoc who worked on the project.

A fortuitous discovery

The team made an unexpected discovery in the lab while they were exploring potential new models for trophectoderm, one of the key lineages of the blastocyst. They observed an unusual developmental phenotype during experiments, and traced the trigger for this phenotype to a specific small molecule.

“The reagent dimethyl sulfoxide (DMSO) is widely used as a solvent to dissolve a broad range of chemicals and as a cryoprotectant cell preservation. We noticed that it appeared to be doing something far more profound in our setup, so we followed the science,” says Li.

“Existing trophectoderm and blastoid models typically rely on combinations of multiple signaling molecules,” says Samhan Alsolami, a former KAUST team member, now at the Salk Institute for Biological Studies in California, United States. Alsolami remains affiliated with KAUST via the Ibn Rushd fellowship program. “We found that DMSO alone could induce trophectoderm-like differentiation. Building on this, we showed that DMSO is sufficient to drive robust blastoid formation from human pluripotent stem cells without the need for any additional factors,” he explains.

The simplicity of the model surprised the team, who were expecting to need a complex cocktail of molecules and signals to generate functioning blastoids.

A first for embryogenesis research

Using this model, the researchers uncovered a crucial biological mechanism that helps the embryo correctly form its first cavity. They found that tiny structures called lysosomes, located inside cells, must function properly for this cavity to form. The team proved that the molecular pump V-ATPase plays a central role in driving and regulating blastoid cavitation. Furthermore, DMSO treatment upregulates V-ATPase activity to ensure the cavity forms correctly.

“This discovery opens the door to numerous further investigations, including examining why early developmental failure happens in some embryos but not others,” says Yiqing Jin from the KAUST team. “Our model may help researchers understand early implantation-related defects, pregnancy loss, and developmental abnormalities that begin before pregnancy is even clinically recognized.”

“The model could also be used to investigate congenital disease mechanisms in a controlled setting,” adds Alsolami. “The system’s simplicity and reproducibility make it well suited for mechanistic studies and screening approaches.”

Future goals

The KAUST team is extending this work to build models that capture all stages of embryogenesis with greater fidelity. In particular, Chandrasekaran is developing a model that captures the features of day 14 of human embryogenesis.

“Our vision is not only to generate embryo models but to use them as discovery platforms,” concludes Li. “We hope to combine stem cell biology with imaging, genomics, and functional perturbation approaches to answer deeper questions about how mammalian embryos self-organize, why development sometimes fails, and how these earliest events influence health later in life.”

 

Modern biology is awash in data. Scientists can sequence DNA, track gene activity cell-by-cell, map proteins in space, and image tissues at microscopic resolution. However, it is a struggle to put all that information together to form a cohesive view.

A KAUST-led vision for artificial intelligence (AI) could help bridge that gap. An AI system that combines multiple biological data modalities into a single model has been described by members of the AI4BioMedicine lab in the Biomedical Division. Called a “super transformer”, the new AI architecture aims to turn today’s fragmented measurements by different technologies into a more coherent picture of life inside cells and tissues^[1].

“This bridges the gaps between siloed computational approaches,” says Jesper Tegnér, professor of bioscience and computer science at KAUST who led the work.

“Such integration will be necessary if AI is to move beyond narrow, single-purpose biological analyses,” says research scientist Sumeer Khan, who co-authored the paper with Xabier Martínez de Morentin, a postdoctoral researcher. The proposed architecture, he explains, is meant to “facilitate scalable integration across data types” and thus provide a framework that can be used across genomic and biomedical research.

The idea fits into a broader effort by Tegnér and his KAUST colleagues to build AI systems that can both integrate biological measurements and explain what they have inferred from them. A single model that can learn from DNA sequences, gene activity, tissue images, and other data simultaneously could begin to link cause and effect across levels of biology, connecting genetic changes to altered cells, tissues, and, eventually, disease.

The appeal of such a system becomes clearer when considering how biological data are analyzed today. Most computational tools are built for a single task: one algorithm for DNA sequences, another for single-cell gene expression, another for tissue images. Integrating their outputs often requires bespoke pipelines, expert judgment, and guesswork. And as datasets grow in scale and complexity, this patchwork approach begins to fray.

Transformers offer one possible way forward. Originally developed for language processing, they were designed to understand how words relate to one another across sentences, paragraphs or extended treatises.

Their key innovation lies in a machine-learning technique called “attention,” which enables models to weigh relationships across a dataset. Rather than processing information strictly in order, a transformer learns which elements matter most to one another, even when they are far apart. That ability proved essential for systems that translate languages or summarize documents, and the same logic applies to biology.

Biological systems are full of distant and indirect interactions. Genes influence one another across long stretches of DNA. Cells respond to signals from their neighbors and from faraway tissues. Molecular events, including disease-related deficiencies, can ripple upward to shape organs and whole organisms. In that sense, life — and disease — have a grammar of their own, and transformers trained on enough data may be able to learn it.

The “super” in “super transformer” reflects an ambition to extend this approach. Rather than applying transformers to a single data type at a time, the KAUST team envisions an architecture that can handle multiple modalities simultaneously. In their proposal, DNA sequences, gene-expression profiles, spatial maps and images would all be translated into a shared internal representation, then linked through the same attention-based machinery.

That vision comes with important caveats. As Tegnér and his collaborators reported late last year, architectural choices and the fine structure of large neural networks can strongly influence robustness, bias, and interpretability, particularly when biological data are noisy or incomplete^[2]. Scaling models without careful design can amplify spurious correlations rather than reveal meaningful structure.

Those concerns motivate the design of the next generation of AI systems tailored for biomedicine — with innovations emerging from across KAUST laboratories. Taken together, these efforts point toward a future in which AI acts less like a collection of specialized tools and more like a unifying layer for biology. It could integrate diverse data, reason across them, and return answers that make sense to human researchers.

In that light, the “super transformer” is best understood not as a finished product, but as a blueprint for how biological AI might finally be built to connect scales rather than fragment them.

Many practical challenges remain, from data standardization to computational cost. Still, Tegnér argues that the direction is clear. Biology no longer lacks data; it lacks a way to see the whole.

 

Electronic devices based on gallium oxide can operate at temperatures even colder than deep space, KAUST researchers have found^[1]. This capability could eventually support extreme-temperature applications such as quantum computing and space exploration.

Computer chips, sensors, and other electronic systems all depend on semiconductors. These materials have an energy gap, known as the band gap, that electrons must leap to conduct electricity. At low temperatures, however, electrons become trapped and cannot move, a phenomenon known as freeze-out.

“In practice, most conventional electronics start to fail as you go below about 100 K (−173 °C),” says Vishal Khandelwal, a former Ph.D. student in Xiaohang Li’s group, who led experimental work on the new devices.

Since electronics are exposed to far colder temperatures in space, or in quantum computers that run at just 4 K, they require thermal management systems that add cost, bulk, and complexity.

The KAUST team has a long history of research on the ultrawide-bandgap semiconductor beta-gallium oxide (β-Ga2O3), previously demonstrating its resistance to radiation and high temperatures. Its wide bandgap means that devices based on gallium oxide experience less current leakage and keep working even at 500 °C, far beyond the capabilities of ordinary silicon circuits.

Earlier studies also showed that the material does not suffer from the freeze-out effects of other semiconductors. To exploit that effect, the researchers have built two devices based on beta-gallium oxide seeded with silicon atoms. This additive, known as a dopant, supplies electrons that help current to flow in the devices.

The first device is a fin field-effect transistor (FinFET), featuring fin-shaped channels that make it stronger and more stable than conventional field-effect transistors. The second is a logic component called an inverter (also known as a NOT gate), a fundamental building block of computer circuits. Both devices demonstrated reliable performance at just 2 K.

At that temperature, there is almost no thermal energy to help electrons jump into gallium oxide’s conduction band. “Instead, the electrons hop through an ‘impurity band’ created by the silicon atoms, enabling the device to carry a current,” Li explains.

Although these are not the first electronic devices to operate at 2 K, this is the first demonstration of an ultrawide-bandgap semiconductor used to build transistors and logic inverters at such low temperatures. “Practically speaking, it allows the development of compact cryogenic circuits made from one material,” says Li — potentially simplifying electronics in quantum computers.

“But its greatest advantage may be for space applications,” he adds. “Space probes face huge temperature swings, so devices that work from a few K to hundreds of K — like beta-gallium oxide — could reduce the need for bulky thermal protection.”

The researchers plan to use beta-gallium oxide to build a toolbox of other devices, including radio-frequency transistors, photodetectors, and memory cells. “We have demonstrated the basic building blocks,” says Li. “Now the work is to scale this up into complex cryogenic chips and to push the limits of performance in this ultracold regime.”

 

Reliable electricity supply is vital in desert locations, where maintaining cooling systems during heatwaves can be essential for human health. For communities considering a shift to renewable energy, accounting for extreme weather events can help prevent electricity shortfalls, KAUST researchers have shown^[1].

Recent advances in renewable energy generation and storage are leading many sustainability-focused communities, including the KAUST campus, to explore how they might transition from fossil-fueled electricity to local grids powered entirely by renewable energy. “These systems must be carefully designed to ensure reliability,” says Farah Souayfane, a research scientist in Omar Knio’s lab, who led the work.

“Most existing designs for community-scale renewable energy systems in hot desert regions like Saudi Arabia optimize performance for average weather conditions,” Souayfane explains. “This approach could lead to failures during rare but critical weather events,” she adds.

Extreme weather days — characterized by very hot, calm, and cloudy conditions — combine high electricity demand for cooling with low electricity supply from wind and solar. This mismatch wouldcan result in power failures in  systems not designed for such conditions.

“We sought to explicitly account for extreme weather in renewable energy systems designed for hot desert communities and quantify the cost implications by designing a resilient renewable energy system for KAUST,” says Ricardo Lima, a research scientist in Knio’s group.

The team based their analysis on a 25-year historical record of hourly weather data for KAUST’s location. “The system was first optimized for a single year of data, then simulated over the full 25-year period to identify failure events when supply did not meet demand,” Souayfane says. These extreme conditions were progressively incorporated into the design, with additional electricity storage and generation capacity added until the system could reliably meet energy demand.

“The system balances cost and resilience by combining concentrated solar power, photovoltaic panels and wind turbines with battery and thermal storage,” Lima says. “Resilience was further improved by using the KAUST desalination plant’s flexible energy demand to reduce stress on the system during extreme events.”

The team found that the optimized system could reliably meet KAUST’s electricity demand during historical extreme conditions while avoiding more than 330,000 tonnes of CO₂ emissions annually compared with fossil fuel electricity supply. “Achieving this level of reliability requires additional investment, increasing system costs by 19 to 30 percent depending on configuration,” Lima notes.

The analysis provides KAUST with a practical framework for designing a resilient, low-carbon power system suited to campus-scale applications, Knio says. He adds that for Saudi Arabia, it offers insights into how renewable energy systems can support energy diversification and emissions reduction under harsh climatic conditions.

Next, the team is exploring additional demand-side flexibility options to manage energy usage during extreme events, including district cooling operation and storage flexibility. The researchers are also integrating climate projections to account for future risks as well as historical extremes. “This will support long-term planning and improve resilience metrics for renewable energy systems,” Knio says.

The genes that could help the world’s crops survive drought, heat and disease probably already exist. But much of this genetic diversity remains hidden within ancient plant varieties and forgotten seed collections, among millions of DNA differences that are difficult to spot.

Now, a new way of reading crop genomes is helping scientists uncover these variations.

Instead of comparing plant DNA to a single reference genome, researchers are beginning to scan genomes as collections of tiny fragments known as k-mers. These short strings of genetic code, tipically a few dozen DNA bases long, act like molecular barcodes, allowing scientists to quickly identify which fragments appear in which plants. That makes it possible to compare genetic variation across thousands of samples simultaneously^[1].

The approach is opening a new window into the vast genetic diversity within corn, rice and other major food crops that feed billions of people but face growing pressure from climate change. Scientists at KAUST are at the forefront of this work.

From large-scale genomic surveys of bread wheat and its wild relatives to new strategies for mining diversity in global seed banks, researchers are showing how k-mer–based analyses can identify rare genetic differences that were lost as crops were bred into modern varieties. Because the method can scan thousands of plants at once, it helps quickly pinpoint where useful traits exist across large seed collections.

These tools are beginning to transform seed banks from static archives into dynamic research resources, helping breeders identify plants that carry valuable traits.

“Representing genomes as collections of k-mers provides a scalable way to capture and compare genomic diversity across large datasets,” says Simon Krattinger, associate professor of plant science at KAUST.

“The goal is to rapidly identify, test, and introduce genetic diversity for specific genes into crop improvement pipelines,” he adds. “This represents a significant change in how breeders and researchers will use gene banks in the future.”

Many KAUST investigators, including Ikram Blilou, Jesse Poland, Rod Wing, and Brande Wulff, all faculty members of the Plant Science program in the Biological and Environmental Science and Engineering Division, are using k-mer approaches to study the genomes of crops and desert-adapted plants, from date palm and pomegranate to thorn jujube and other species native to Saudi Arabia. Few plants, however, demonstrate the value of k-mer genomics as clearly as wheat, one of the most widely cultivated crops in the world.

In 2024, Krattinger and Wulff spearheaded the development of a new genomic resource for Tausch’s goatgrass (Aegilops tauschii), a wild grass from which part of modern bread wheat’s complicated genome originally came^[2].

The researchers sequenced the genomes of dozens of goatgrass plants collected from across the species’ natural range, from Turkey in the west to China in the east. They then broke each genome into millions of k-mers and combined these overlapping DNA fragments into a large searchable database. By comparing goatgrass genomes with modern wheat varieties, the team identified how much genetic diversity had been lost during thousands of years of domestication.

“Those missing genes represent the raw material for selection and breeding,” Krattinger explains. “Reintroducing them is a key priority for crop improvement,” he adds.

As a proof of concept, the researchers focused on a goatgrass gene that protects against a destructive fungal pathogen. According to Krattinger, this disease-resistance gene offers a promising starting point for breeders looking to develop crops that can better withstand fungal infections, a major cause of yield losses in wheat harvests worldwide.

More recently, Krattinger and his former Ph.D. student, Emile Cavalet-Giorsa, used k-mer–based methods to trace the origin of a key genetic changes that enabled wheat domestication: grain retention^[3].

In wild cereal plants, mature seed heads typically shatter, allowing grains to fall and disperse — a process that supports natural reproduction but complicates harvesting. Early farmers therefore selected plants whose seeds remained attached to the stalk.

The k-mer–based analyses showed that this trait did not result from a single mutation that farmers selected, as previously believed. Instead, multiple mutations responsible for keeping wheat grains attached appeared in wild wheat populations tens of thousands of years before agriculture began. These pre-existing variants, scattered across different wild populations, were likely later combined and selected by early farmers.

The findings reinforce a central lesson of the KAUST team’s work: many of the traits needed for future crop improvement may already exist in seed banks and wild plants. With continued advances in k-mer genomics, these vast collections may become searchable maps of crop diversity and a an important resource for crop breeding.