Tens of thousands of years ago, as modern humans migrated out of Africa and into unfamiliar territories, they encountered Neanderthals—a now-extinct group of ancient humans who lived in Eurasia. These interactions left a lasting mark on our DNA; today, nearly all non-African humans carry traces of Neanderthal DNA, a genetic inheritance that provides important insights into human migration and survival.

Despite years of research, however, questions remain regarding the timing, extent, and impact of the genetic exchange between Neanderthals and modern humans.

In two groundbreaking new studies published in Science and Nature, researchers from the University of Rochester—along with colleagues at the Max Planck Institute for Evolutionary Anthropology, the University of California, Berkeley, and others—traced how ancient interactions with Neanderthals shaped modern human evolution. By examining patterns of Neanderthal DNA in both modern and ancient human genomes, they reconstructed a timeline of interbreeding and its evolutionary impacts. Their findings reveal when and where these exchanges occurred—and highlight how Neanderthal genes helped humans adapt to new environments.

“Our study provides the most detailed insights yet into how Neanderthal gene flow impacted human genomes,” says Benjamin Peter, an assistant professor in the University’s Department of Biology. “It helps us understand when Neanderthals and humans interacted, which Neanderthal genes were beneficial for our ancestors, and the forces that influenced genetic diversity and shaped the course of human evolution.”

Human-Neanderthal divergence and gene flow

About 500,000 years ago, Neanderthals and modern humans diverged from a common ancestor. Following this split, one lineage evolved into Neanderthals in Eurasia, while the other evolved into modern humans in Africa. Both groups had complex behaviors, including tool use and social structures, but Neanderthals developed unique physical traits, such as a stockier build and larger brow ridges, reflecting adaptations to their environment.

Sometime around 40,000 to 60,000 years ago, modern humans left Africa, encountering Neanderthals and interbreeding. The gene flow between Neanderthals and modern humans resulted in most non-Africans carrying one to two percent Neanderthal DNA.

However, the exact timing of the gene flow event has remained elusive.

A single period of gene flow

To uncover a more precise timeline, the researchers used genome sequencing techniques to analyze more than 300 genomes from ancient and modern humans spanning the last 50,000 years. They examined Neanderthal DNA segments in the individual genomes across different time periods and geographic regions, identifying patterns to determine when interbreeding occurred and how natural selection influenced which Neanderthal genes were passed on.

They found that most Neanderthal DNA in modern humans can be traced to a single major period of gene flow, which occurred about 47,000 years ago and lasted approximately 7,000 years. This suggests there was one extended period of interaction between the two groups, rather than multiple separate events, as some researchers had previously believed.

The findings provide tighter constraints for when humans migrated out of Africa—known as the “Out-of-Africa event”—helping to more precisely pinpoint when migration and interbreeding occurred.

“Our results suggest that all Neanderthal ancestry in living people traces back to the same event shortly after the Out-of-Africa event,” Peter says.

Ancient traits stand the test of time

So, what Neanderthal genes are in humans? The researchers found that Neanderthal DNA is not evenly spread throughout the genome. In fact, some regions lack Neanderthal DNA entirely, suggesting that Neanderthal ancestry in those areas wasn’t beneficial for survival. Other regions—particularly those linked to traits such as skin pigmentation, metabolism, and immune function—have higher concentrations of Neanderthal DNA. The researchers discovered that this uneven distribution existed in human genomes more than 40,000 years ago, indicating that some Neanderthal genes provided immediate benefits, such as helping humans adapt to new climates as they migrated out of Africa.

“This shows that natural selection on Neanderthal genetic variants, both beneficial and harmful, acted very rapidly and was likely quite strong,” Peter says.

The patterns in the DNA also suggest that interbreeding may not have been entirely random. Instead, factors such as geography or culture may have influenced which groups of humans were more likely to interact with Neanderthals, leading to different amounts of Neanderthal DNA in different populations.

Ancient encounters and future discoveries

Uncovering when and how certain Neanderthal genes were passed down reveals how these ancient encounters shaped human adaptation and diversity.

These latest findings not only show the lasting influence of Neanderthal genes on human evolution, but they also pave the way for future research. Gathering more genetic data from areas of the world where early human-Neanderthal interactions remain a mystery could deepen the understanding of this key moment in human history.

“A major limitation of our study is that we do not have genetic data from early modern humans in several key areas of the world, such as the Middle East, South Asia, and Oceania,” Peter says. “Direct data from there would likely allow more insights on where exactly Neanderthals and humans met.”

Beneath the ocean’s surface, simple marine animals called sea sponges grow delicate glass skeletons that are as intricate as they are strong. These natural structures are made of a material called silica—also known as bioglass—that is both lightweight and incredibly durable, allowing the sea sponges to thrive in harsh marine environments.

Now, scientists at the University of Rochester have replicated this remarkable material in the lab, using bacteria and enzymes from sea sponges to create tiny microlenses that mimic the sea sponge’s natural ability to combine strength and lightness. In a paper published in the journal PNAS, the team—including scientists from the University of Colorado–Boulder, Delft University of Technology, and Leiden University—reports that the bioinspired material could pave the way toward specialized image sensors for medical and commercial uses. By applying the remarkable properties of sea sponges, the researchers unlock new possibilities for creating sustainable and efficient materials that mimic the natural world.

“This research is the first to engineer light-focusing properties into bacteria cells, and I am excited to explore the different possibilities that our work has opened up,” says Anne S. Meyer, an associate professor in Rochester’s Department of Biology.

What is a microlens?

A microlens is a very small lens that is only a few micrometers in size—about the size of a single cell in your body. Microlenses are designed to capture and focus or manipulate light into intense beams at a microscopic scale.

Because of their small size, microlenses are typically difficult to create, requiring complex, expensive machinery and extreme temperatures or pressures to shape them accurately and achieve the desired optical effects.

When Meyer learned about the enzymes that sea sponges use to make their glass skeletons—and that the glass structures had excellent optical properties—“it seemed like a perfect basis for a synthetic biology project,” she says.

Collaborative innovation across disciplines

Meyer teamed up with experts across multiple disciplines, including optics, physics, and chemistry. Her lab engineered bacteria cells to express the silicatein enzyme from sea sponges, which the animals use to mineralize silica-based glass. They also developed a novel microscopy technique to measure the optical properties of the bacteria cells. In collaboration with material scientists at the University of Colorado–Boulder, Meyer ensured that silica was present on the engineered cells by analyzing the bacteria’s chemical properties. She also worked with faculty members Greg Schmidt and Scott Carney at Rochester’s Institute of Optics to create mathematical models that predicted the optical properties of the glass-coated cells.

The result? Bacterial microlenses that are much smaller than typically produced microlenses.

Because the microlenses are created by bacterial cell factories, they are inexpensive and easy to grow, and they can create their glass coating at standard temperatures and pressures.

“These properties make them well-suited for a unique range of applications,” Meyer says.

Small lenses, big potential

What are the benefits of a microlens? The tiny size of the bacteria-based microlenses makes them ideal for creating higher-resolution image sensors that go beyond current capabilities. The microlenses could, for instance, allow clinicians to visualize smaller structures with greater clarity. Since the glass-coated bacteria focus light into very bright beams, they have the potential to enhance conventional microscopy by enabling the imaging of objects that are currently too small to be visualized, such as small subcellular features.

The glass-coated bacteria remain alive for several months after glass encapsulation, making them living optical devices that could be used to sense and respond to their environment by changing their optical properties.

These characteristics make the microlenses attractive for other environments as well: Meyer and a team of her colleagues recently received a grant from the Air Force Office of Scientific Research to study the effects of the materials in low-gravity environments.

“The ease of producing these microlenses could make them a good way to fabricate optics in locations with less access to nanofabrication tools, including outer space,” Meyer says.

Biologists at the University of Rochester are calling a discovery they made in a mysterious region of the chromosome known as the centromere a potential game-changer in the field of chromosome biology.

“We’re really excited about this work,” says Amanda Larracuente, the Nathaniel and Helen Wisch Professor of Biology, whose lab oversaw the research that led to the findings, which appear in PLOS Biology.

The discovery involves an intricate and seemingly carefully choreographed genetic tug-of-war between elements in the centromere, which is responsible for proper cell division. Instead of storing genes, centromeres anchor proteins that move chromosomes around the cell as it splits. If a centromere fails to function, cells may divide with too few or too many chromosomes.

These critical structures are rich in what biologists call “selfish” genetic material—transposable elements that move within the genome, and thousands of repeated segments of DNA known as “satellite DNAs”—that often compete during cell division to ensure their own transmission.

For centromeres to function effectively, though, these competing elements must also cooperate.

“In biology, we’re used to thinking about things that have essential roles as being highly conserved,” Larracuente says. “So, it’s fascinating that they are the opposite of highly conserved. They are rapidly evolving.”

‘Dramatic’ centromere reorganization

To learn more about the interplay between these elements, researchers studied closely related species of fruit flies, or Drosophila, and found that centromeres frequently switched between types of transposable elements and satellite DNA in short spans of evolutionary time.

“Repetitive sequences are known to evolve rapidly in general,” Larracuente says. “But what we found was a dramatic centromere reorganization over two short evolutionary timescales.

“We didn’t just see different variants of the same sequence in different species, we found categorical shifts in the types of elements.”

The researchers used chromatin profiling and high-resolution imaging on stretched chromatin fibers to observe these shifts in detail.

“Regardless of the evolutionary forces driving this turnover,” reads the study, “the rapid reorganization of centromeric sequences over short evolutionary timescales highlights their potential as hotspots for evolutionary innovation.”

The lab is interested in understanding the roles these DNA sequences play in centromere function and stability in future work. Larracuente says the discovery and subsequent study of centromere dynamics could have potential applications in the long-term for how we treat diseases and disorders characterized by genome instability, such as cancer, and other aging-related diseases.

“Those can be related to centromere defects,” Larracuente says. “Learning how DNA sequences contribute to centromere organization and function could help us understand abnormal centromere behavior.”

Imagine being able to inscribe a new pattern of activity into a person’s brain that would allow for faster learning, or better treatment of psychiatric and developmental disorders such as depression or autism. Now imagine being able to do that in a way that doesn’t require brain surgery or any physical manipulation. Sounds like science fiction?

It still is. But that’s exactly what Coraline Iordan, an assistant professor of brain and cognitive sciences and of neuroscience at the University of Rochester has been working toward, showing for the first time that it can certainly be done for learning new visual categories of objects.

Generally, learning happens when our brain changes through experience, study, or instruction. But Iordan and colleagues at Yale and Princeton successfully tested a novel approach for teaching the human brain to learn through external manipulation and neural feedback—what they call the “sculpting” of brain activity patterns. The research appears in the Proceedings of the National Academy of Sciences.

“With our method not only can we nudge complex patterns around in the brain toward known ones, but also—for the first time—write directly a new pattern into the brain and measure what effect that has on a person’s behavior,” says lead author Iordan.

Brain sculpting—a new approach to learning?

The scientists used real-time neuroimaging and second-by-second neurofeedback to modify how the brain represents and processes information about visual objects. Lying inside a functional magnetic resonance imaging (fMRI) machine, study participants viewed objects projected onto a mirror above their heads, which looked like a small screen. The object_­—an abstract shape that some participants described as a petal, plant bulb, or butterfly—pulsed gently on the participants’ mirror until they managed to “move it” by their own thought processes to the pattern of activity in their brain (monitored via fMRI in real time) that the scientists had previously chosen. The researchers instructed the participants to “generate a mental state” that would reduce the shape’s oscillation but had not taught the study participants how to achieve such mental state.

“One of the striking features of the study is that the neural responses and corresponding behavior to the new categories occurred without explicit awareness of those categories, showing that a long tradition of work in psychology on implicit processing—that is, the ability to respond to information meaningfully outside of awareness—also extends to the learning and formation of new neural representations,” says coauthor Jonathan Cohen, a cognitive neuroscientist at Princeton University.

Researchers are developing a new system that will allow people to speak anonymously in real time through computer-generated voices to help protect privacy and avoid censorship or retaliation. The technology is intended to help people such as intelligence officers carrying out sensitive missions, crime witnesses concerned about being identified by perpetrators, and whistleblowers who fear retaliation.

The three-year project, led by Honeywell and including collaborators from the University of Rochester, Texas A&M, and the University of Texas at Dallas, is funded by the Intelligence Advanced Research Projects Activity (IARPA) and part of the Anonymous Real-Time Speech (ARTS) program.

The voice-changer project has three main objectives. First, the system will transform what a user says into a digital voice within a few milliseconds, ensuring that it can be used in real-time conversations. Second, the team aims to allow users to specify what they call static traits, allowing control over the digital voice’s age, gender, and dialect. Lastly, they want to neutralize what they call dynamic traits, such as emotions or health status that could potentially tip the identity of the user.

“In the end, a 30-year-old woman from Texas will be able to instantaneously transform her voice to be output by the virtual speaker to sound like a 50-year-old man with a British accent, for example, without producing artifacts that can be traced back to the identity of the user,” says Zhiyao Duan, an associate professor of electrical and computer engineering and Rochester’s lead on the project. “And in addition to the latency requirements, we’ll also be working to ensure the intelligibility and naturalness of the computer-generated voice.”

Duan says that while the roles on the project are fluid, his team at Rochester will initially focus on generating the virtual speakers and controls for the static traits, building on their experience in speaker modeling, disentangled speech representations, and voice synthesis. The team will first develop the technology to work in English. If successful, they plan to expand it to other languages such as Spanish, Mandarin, and Korean.

The team hopes these open-source voice-changer tools will have positive benefits far beyond the intended initial use cases. Still, the researchers recognize that people may have concerns about such powerful software.

“I think it’s natural for people to wonder what will happen if these tools get in the hands of bad actors,” says Duan. “It’s important to note that my lab and others around the globe are also working to develop deepfake detection tools so that people can discern whether something is said by an actual person or generated through algorithms. Those tools will be equally important to have.”

Rising carbon dioxide levels are exacerbating the effects of climate change, highlighting the need for solutions that balance energy demands with environmental sustainability.

What if there were a way to turn carbon dioxide itself into a resource—producing energy while reducing harmful emissions at the same time?

A team of 11 undergraduate students at the University of Rochester has done just that. The team created a carbon-negative energy source that uses bacteria to generate energy while simultaneously capturing and storing carbon dioxide. Their innovative approach not only addresses energy needs but also produces ethanol as a sustainable biofuel.

In October, the team—called Team CyanoVolt—submitted their research to the 2024 International Genetically Engineered Machine (iGEM) competition, where student-led teams from around the globe compete to tackle real-world challenges using synthetic biology. Synthetic biology leverages engineering principles to create biological components inspired by nature.

The Rochester team competed against more than 400 teams from around the world and was awarded a gold medal for their project.

“The idea of a carbon-negative energy technology was exciting to the team because it can clean the atmosphere while also creating clean energy,” says Anne S. Meyer, an associate professor in the Department of Biology and one of the advisors of Rochester’s iGEM team. “I was incredibly impressed by the students’ smart project design and their dedication to keep trying to get their experiments to work over many, many rounds of optimization and improvement.”

Harnessing energy from bacteria

As part of iGEM, undergraduate students from a variety of majors design and build engineered biological systems using DNA technologies. Students lead every aspect of the project, including selecting the topic, conducting experiments, managing budgets, raising funds, and handling social media.

This year’s iGEM team began brainstorming project ideas in the spring, spent the summer and early fall developing and conducting experiments, and submitted their project for judging in the iGEM competition in late fall. The idea for Team CyanoVolt’s project was inspired by a challenge posed by one of the team member’s high school teachers.

“Our teacher challenged us to think of ways to reuse food waste,” recalls Grace Widjaja ’26, a biochemistry and music double major. The task opened Widjaja’s eyes to the idea of finding power sources in unexpected places.

Later, when she joined iGEM in her sophomore year of college, she remembered that challenge and thought, “If we can get energy from food, why not from other sources, like bacteria?”

Building on this idea, the Rochester team settled on a project that combined three key goals: capturing carbon dioxide, producing electricity, and generating ethanol, by genetically engineering bacteria and designing specialized biophotovoltaic cells.

“By accomplishing these three goals, we hoped to propose a solution to alleviate our climate crisis,” Widjaja says. “I learned about synthetic biology for the first time in my freshman year of college. I had always heard of GMOs as being these terrible things, but iGEM showed me how to find ways to genetically modify organisms to produce products that help the world.”

The power of photosynthesis

To capture carbon dioxide, the team genetically engineered cyanobacteria—a type of bacteria that naturally carries out photosynthesis—to optimize the bacteria’s ability to absorb light and carbon dioxide. The modified bacteria operate like a tiny energy factory: when exposed to light, the bacteria capture carbon dioxide and transform it into energy-rich molecules. However, instead of just storing this energy for their own use, the engineered cyanobacteria release some of the energy as free electrons.

To harness the energy from the electrons, the team designed specialized biophotovoltaic cells. These devices function like solar cells but generate electricity using biological materials and processes—in this case, the cyanobacteria undergoing photosynthesis. The biophotovoltaic cells capture electrons and direct them into a circuit, producing electricity.

The team also developed a novel screen printer that prints customized structures within the biophotovoltaic cells to ensure the cyanobacteria are optimally placed and distributed, as well as to facilitate efficient flow of electrons.

Fueling the future

The team took the project a step further by genetically engineering the cyanobacteria to produce ethanol—a useful biofuel—as a byproduct while they carry out photosynthesis. The unique process of producing both power and biofuel simultaneously makes Team CyanoVolt’s project both environmentally friendly and efficient. The result is a renewable system that is carbon-negative; instead of just reducing emissions, it actively decreases the total amount of carbon dioxide in the environment by absorbing the carbon dioxide from the atmosphere during photosynthesis.

“It was really novel to see bacteria being used in this way and to understand that cyanobacteria has much more capability than is often discussed in literature,” says Seeya Khattar ’26, a molecular genetics major. “As the bacteria generate power, they help lower greenhouse gas levels, offsetting any emissions they might produce and ultimately creating a net reduction in atmospheric carbon dioxide.”

Powering through challenges

Although the team faced challenges—most notably, they initially had a difficult time growing cyanobacteria in the lab—iGEM offered a unique experience to not only lead a research project from start to finish but to solve problems that led to meaningful results.

“In research, there are a lot of problems that come up,” says Claire English ’26, a computational biology and statistics double major. “We had the hardest time growing our cyanobacteria, and now we have all these samples that are super vibrant and green and it’s the most exciting thing. Sometimes in research you question whether this is what you should be doing, but then there are these experiences, like when we finally made electricity for the first time, that are just so rewarding.”

Team CyanoVolt’s project is documented on its Wiki page, offering a resource for future students or developers to refine and expand upon the ideas.

Can artificial intelligence-powered tools help enrich child development and learning?

That question is the crux of a series of research projects led by Zhen Bai, an assistant professor of computer science at the University of Rochester and the Biggar Family Fellow in Data Science at the Goergen Institute for Data Science. From tools to help parents of deaf and hard-of-hearing (DHH) children learn American Sign Language (ASL) to interactive games that demystify machine learning, Bai aims to help children benefit from AI and understand how it is impacting them.

Bai, an expert in human-computer interaction, believes that, despite all the concern and angst about AI, the technology has tremendous potential for good. She believes children are especially primed to benefit.

“Over the years, I’ve seen how kids get interested whenever we present technology like a conversational agent,” says Bai. “I feel like it would be a missed opportunity if we don’t prepare the next generation to know more about AI so they can feel empowered in using the technology and are informed about the ethical issues surrounding it.”

Minimizing language deprivation in deaf and hard-of-hearing children

During one of Bai’s earliest experiences at the University, she met a key collaborator who led her to a new avenue of research. At a new faculty orientation breakfast, she happened to sit next to Wyatte Hall, a Deaf researcher and assistant professor at the University of Rochester Medical Center’s Department of Public Health Services. The two bonded over a shared interest in childhood development and learning.

Hall explained some of the unique challenges children who are deaf and hard-of-hearing face in cognitive and social development. More than 90 percent of DHH children are born to hearing parents, and often the very first deaf person that parents meet is their own baby. In early human development, there’s a neurocritical period of language acquisition—approximately the first five years of a child’s life—in which children need to acquire a first language foundation. Having parents who do not know a signed language, and the limits of technology such as the cochlear implant and hearing aids, increases the risk of DHH children experiencing negative developmental outcomes associated with language deprivation.

“I learned a lot from Dr. Hall about this concept of language deprivation and became fascinated with the idea of how technology could play a role to make life easier,” says Bai. “I wanted to explore how to help facilitate this very intimate bonding from day one between parents and their kids.”

Bai and Hall began collaborating on a project called the Tabletop Interactive Play System (TIPS) to help parents learn ASL in a natural setting. The system uses a camera and microphone to observe the parent and child interacting, and then uses a projector to present videos of relevant signs retrieved via artificial intelligence from multiple ASL libraries.

In addition to a tabletop version, Bai has been developing versions for tablets, smart watches, and smart glasses, together with her team of undergraduate and graduate students with backgrounds in computer science, data science, and neuroscience. She has also collaborated with student fellows from the Rochester Bridges to the Doctorate program and other researchers from the Deaf community such as Athena Willis, a scholar in the Rochester Postdoctoral Partnership from the University’s Department of Neuroscience.

Rochester, reportedly home to the country’s largest population of DHH people per capita, is a uniquely rich setting for researching assistive technologies for the Deaf community. Hall says Bai’s willingness to learn from and collaborate with the Deaf community has helped improve the effectiveness of the tool.

“Often we’ve seen hearing people, hearing researchers become involved in Deaf-related things, they learn something interesting about Deaf people and want to run with it for their own work. Even with the best of intentions, that can go awry very quickly if they are not collaborating with Deaf people and the community at all or in the right way,” says Hall. “My experience with Dr. Bai, though, she really started with a good foundation and kept collaborating with me in a very positive way, so it’s been a great partnership from the very beginning.”

Demystifying machine learning

As AI provides more recommendations to kids about the books they read, shows they watch, or toys they buy, Bai wants to provide learning opportunities so kids can use the technology and understand how it works to make it less of a “black box.” She earned a prestigious Faculty Early Career Development (CAREER) award from the National Science Foundation to develop technologies that help K–12 students demystify machine learning, an integral aspect of current approaches to AI.

Partnering with researchers from the Department of Computer Science, including Albert Arendt Hopeman Professor Jiebo Luo, and from the Warner School of Education—including Frederica Warner Professor Raffaella Borasi, Associate Professor Michael Daley, and Associate Professor April Luehmann—her team developed visualization tools that help K–12 students and their teachers use machine learning to make sense of data and pursue scientific discovery, even if they do not have programming skills.

Bai has been piloting the web-based tool her team developed, GroupIt, with K–12 teachers to see how she can help the next generation make sense of big data. She says working with teachers has been crucial because they are on the frontlines of helping children make sense of AI.

“Teachers play such a critical role in integrating AI education in the STEM classroom, but it’s so new for them both technologically and pedagogically,” says Bai. “We want to empower teachers with easy-to-use tools so they can create more authentic learning activities that integrate data into their classroom, whether they’re teaching hard sciences or social sciences.”

To help K–12 students understand how AI is affecting them, Bai and her students also developed an augmented reality game. The game uses bee-pollinating flowers as an analogy for AI-powered recommendation systems, illustrating how the preference selection process works. Called BeeTrap, the game shows how choosing to pollinate certain types of flowers can reduce the overall biodiversity of the flowers in the environment.

“BeeTrap explains the mechanism that makes recommendations more or less relevant and diverse to a person,” says Bai. “The goal is to help children realize the value of information and how things are being selectively recommended to people based on previous choices they made and other personal information.”

Bai says this is especially important for marginalized groups, who can be impacted by inherent biases in AI systems related to race, ethnicity, gender, and other factors. Bai has introduced the BeeTrap game to students in various summer camps including the Upward Bound pre-college program by the David T. Kearns Center at the University, and the Freedom Scholars Learning Center in the city of Rochester.

The team is also creating more tangible representations of AI. Her group created OptiDot, a 3D-printed optical device that shows how AI might suggest different food choices based on your preference for sweet or salty snacks or fatty or healthy options.

Ultimately, Bai thinks it will take a multifaceted approach to help students harness the power of AI, but she is excited to develop tools that can help them get there.

“There is a lot more work to be done to improve the learning experiences and make AI accessible and relatable for students,” says Bai. “We’re happy to play a role in helping to make that happen.”

There are billions of tiny packets of cellular material called extracellular vesicles (EVs) that are produced by cells and released into each person’s blood, saliva, and other bodily fluids. EVs contain invaluable information, such as proteins and genetic material, from their original cell, which can provide insight about the status of the body. Scientists have been trying to leverage EVs for their diagnostic and therapeutic potential but have struggled to do so in a fast and cost-effective way.

In a study published in Small, researchers at the University of Rochester outline a new method for using ultrathin membranes to easily identify EVs for rapid liquid biopsies. The method, called catch and display for liquid biopsy (CAD-LB), holds promise for diagnosing cancer quickly and affordably, and assessing the progress of therapies used to treat diseases.

“By searching samples of blood or other bodily fluids for these extracellular vesicles and the biomarkers they carry, you can find important clues that something is amiss in the body,” says James McGrath, the William R. Kenan Jr. Professor of Biomedical Engineering and leader of the study. “The idea has been around for a while, but previously it required many purification steps to isolate the EVs away from other components of the biofluid. CAD-LB is much simpler and faster, which gives it the potential for clinical use that more complex methods lack.”

The team developed ultrathin membranes with pores sized perfectly to catch EVs. Once a sample of blood is taken, it is quickly processed, injected with a pipette onto a membrane, and directly analyzed under a microscope. By counting the number of pores that glow with the biomarker for the disease being assessed, users can get a quick estimation of how prevalent the disease is within the body.

In addition to outlining the CAD-LB method, the study demonstrated the method’s ability to identify critical immune modulatory proteins on EVs. These proteins play an important role in helping the body fight tumors and can predict how well a patient might respond to immunotherapies.

“CAD-LB is currently sensitive enough to detect some cancers at a curable stage of their development, suggesting the technology’s potential for cancer screening,” says co-author Jonathan Flax, a research assistant professor at the University of Rochester Medical Center’s Department of Urology. “It may also be utilized to predict the patient-specific selection of immunotherapies, the treatment that directs the immune system to target and eliminate cancer cells.”

McGrath credited first author and biomedical engineering PhD student Samuel Walker for his leadership on the study, as well as the Wilmot Cancer Institute and UR Ventures’ Technology Development Fund for providing key financial support.

This presidential cycle has already seen several high-profile examples of people using deepfakes to try to influence voters. Deepfakes are images, audio recordings, or videos generated or modified using artificial intelligence (AI) models to depict real or fictional people. Recent deepfake examples include manipulated audio of Joe Biden urging voters to stay home during primaries and fabricated images of Taylor Swift endorsing Donald Trump.

It appears generative artificial intelligence is an increasingly prominent tool in the misinformation toolbox. Should voters be concerned about being bombarded with phony videos of politicians created with generative AI? An expert in computer vision and deep learning at the University of Rochester says that while the technology is rapidly advancing, deepfake video generation remains harder for bad actors to leverage due to its complex nature.

While OpenAI’s products, including ChatGPT for text generation and DALL-E 3 for image generation, are taking off in popularity, the company has yet to release an equivalent for video generation. According to Chenliang Xu, an associate professor of computer science at Rochester, the company has released previews of its Sora video generation software but has yet to release the product, which is still undergoing testing and refinement.

“Generating video using AI is still an ongoing research topic and a hard problem because it’s what we call multimodal content,” says Xu. “Generating moving videos along with corresponding audio are difficult problems on their own—and aligning them is even harder.”

Xu says that his research group was among the first to use artificial neural networks to generate multimodal video in 2017. They started with tasks like providing an image of a violin player and audio of a violin to generate a moving video of a violin player. From there, they moved on to problems like generating lip movements, and then to creating full talking faces complete with head gestures from a single image.

“Now, we can generate real-time, fully drivable heads and even turn the heads into various styles specified by language descriptions,” says Xu.

Challenges with deepfake detection technology

Xu’s team has also developed technology for deepfake detection. He calls it an area that needs extensive further research, noting that it’s easier to build technology to generate deepfakes than to detect them because of the training data needed to build the generalized deepfake detection models.

“If you want to build a technology that’s able to detect deepfakes, you need to create a database that identifies what are fake images and what are real images,” says Xu. “That labeling requires an additional layer of human involvement that generation does not.”

Another concern, he adds, is making a detector that is generalizable to different types of deepfake generators. “You can make a model that performs well against the techniques you know about, but if someone uses a different model, your detection algorithm will have a hard time capturing that,” he says.

The easiest targets for video deepfakes

Having access to good training data is crucial for creating effective generative AI models. As a result, Xu says politicians and celebrities will be the earliest and easiest targets when video generators become widely available.

“Politicians and celebrities are easier to generate than normal people because there is simply more data about them,” says Xu. “Because so much video of them already exists, these models can use it to learn the expressions they show in different situations, along with their voices, their hair, movements, and emotions.”

But he expects that, at least initially, the training data the “celeb deepfakes” in particular are built on may make them more easily noticeable.

“If you used only high-quality photographs to train a model, it will produce similar results,” says Xu. “It may result in an overly smooth style that you can pick out as a cue to tell it’s a deepfake.”

Other cues can include how natural a person’s reaction seems, whether they can move their heads, and even the number of teeth shown. But image generators have overcome similar early tells—such as creating hands with six fingers—and Xu says enough training data can mitigate these limitations.

He calls on the research community to invest more effort into developing deepfake detection strategies and grappling with the ethical concerns surrounding the development of these technologies.

“Generative models are a tool that in the hands of good people can do good things, but in the hands of bad people can do bad things,” says Xu. “The technology itself isn’t good or bad, but we need to discuss how to prevent these powerful tools from ending up in the wrong hands and used maliciously.”

Scientists are developing advanced tools to understand and treat neurological symptoms such as brain fog associated with respiratory diseases like influenza. The Biomedical Advanced Research and Development Authority (BARDA), part of the Administration for Strategic Preparedness and Response (ASPR) within the US Department of Health and Human Services (HHS), awarded a three-year contract to researchers at the University of Rochester to develop a technology to model respiratory disease effects on the brain and test therapeutic drugs to prevent and treat symptoms. The base-year is funded at $2.4 million with two option years which, if fully funded, would total $7.1 million.

The project will use microphysiological systems (MPS)—small chips with ultrathin membranes supporting 3D networks of human cells, also known as “tissue chips”—to simulate infection and treatment in vitro. This tissue chips will incorporate human lung and brain tissue models.

“This is another step toward making disease modeling and drug discovery focused from the very beginning on more complex, human-relevant systems,” says principal investigator Benjamin Miller, a Dean’s Professor of Dermatology at Rochester with joint appointments in biomedical engineering, biochemistry and biophysics, optics, and materials science. “These chips can help make the whole drug discovery process faster.”

The project builds on work at Rochester’s recently established Translational Center for Barrier Microphysiological Systems (TraCe-bMPS) to build FDA-qualified drug development tools for studying the body’s barrier functions in combating disease. The center was created earlier this year with a $7.5 million grant from the National Institutes of Health.

Co-investigator James McGrath, the William R. Kenan Jr. Professor of Biomedical Engineering and director of TraCe-bMPS, has been using microphysiological systems to study the mechanism by which inflammatory factors can enter the brain through the circulation and cause injury. The new BARDA-funded project will link two of McGrath’s modular, mass-producible chips specialized to mimic different organs.

“This project will connect this ‘brain’ chip upstream of a second chip that models a common source of those injurious factors: the infected lung,” says McGrath. “I’m thrilled to be working with a highly interdisciplinary Rochester team and with BARDA to develop what will be a scientifically important new tool.”

As with long COVID, common viruses such as influenza can produce chronic symptoms such as brain fog, fatigue, and enduring pain. The project offers a new way to explore the relationship between the lungs and brain.

“The respiratory tract, with its cellular, humoral, and hard-wired conduits to the brain, stands as the first line of defense against emerging infectious threats from zoonotic spillovers,” says co-investigator Harris (Handy) Gelbard, director of the Center for Neurotherapeutics Discovery at the University of Rochester Medical Center. “We and our collaborators, with the support of the National Institute on Aging, have worked for the past several years to investigate these mechanisms in the hopes of applying therapeutic agents to ameliorate neurologic disease, especially in the elderly that are vulnerable to these infections. Now, with a world-class team of in-house experts at developing labs-on-a-chip, we have the unique opportunity to fast-track our research in a new lung-to-brain chip.”

David Dean, a professor of pediatrics, biomedical engineering, and pharmacology and physiology, has studied the disease processes that lead to acute respiratory distress syndrome (ARDS) in the hopes of developing new treatments for this devastating disease.

“Studying this required us to use cultured cells from the lung, but almost always, these are grown and studied by themselves, which is not anywhere close to the situation in the lung where over 40 different cell types co-exist and interact to allow us to live. Thus, this is way too simplistic of a model,” says Dean, co-investigator of the project. “On the other extreme, we have used animal models to test hypotheses and drugs in development, but these models are so hard to control and make sense of because so many different things are going on, and it is difficult to attribute a response to a single pathway, leading to a system that is almost too complicated.”

He says the new approach is a win-win solution that will allow the researchers to mimic complex interactions between key cell types in the lung but in a controlled manner.

David Topham, the Marie Curran Wilson and Joseph Chamberlain Wilson Professor of microbiology and immunology and director of the Translational Immunology and Infectious Diseases Institute, will also serve as a co-investigator; Hani Awad, the Donald and Mary Clark Distinguished Professor in Orthopaedics and a professor of biomedical engineering, will act as a consultant. The team will be working with University of Rochester spinout companies Phlotonics to do medium-throughput instrumentation and SIMPore to develop the chips.

The project will last three years, and by the end of the first year, the team aims to link the tissue chip systems with immune cells, demonstrate that they can infect the lung chip with influenza, and observe an inflammatory response in the brain chip. This project has been supported in whole or in part with federal funds from the Department of Health and Human Services; Administration for Strategic Preparedness and Response; Biomedical Advanced Research and Development Authority (BARDA), under contract number 75A50124C00040.

Elected officials voice support for innovative technology development

US Senator Charles Schumer: “I am thrilled to see our researchers leading the charge in groundbreaking medical innovation with this substantial $7.1 million award from the Department of Health and Human Services. This investment speaks volumes about the world-class research happening right here in Rochester. By developing microchips that mimic brain and lung tissue, our scientists are pioneering new ways to understand and combat respiratory diseases and their impact on the brain. This cutting-edge research has the potential to revolutionize our approach to treating these diseases, paving the way for more effective therapies and ultimately saving lives. I remain committed to advocating for robust federal support for scientific advancements that can change the future of healthcare and improve public health outcomes for all.”

US Senator Kirsten Gillibrand: “Researchers at the University of Rochester are leading the charge in disease modeling and drug discovery. This $7.1 million contract from HHS will help researchers at the University of Rochester develop the most advanced technology to model respiratory disease effects and find ways to prevent and treat symptoms. I will continue to fight to secure federal resources to support the innovative work of researchers at the University of Rochester.”

Congressman Joe Morelle: “The University of Rochester continues to drive groundbreaking research, innovation, and scientific advancement in the world of medicine. This significant federal award is further proof of their leadership and limitless potential. I congratulate their team of researchers on their outstanding achievements that will change the way we fight diseases.”