All life is connected in a vast family tree. Every organism exists in relationship to its ancestors, descendants, and cousins, and the path between any two individuals can be traced. The same is true of cells within organisms — each of the trillions of cells in the human body is produced through successive divisions from a fertilized egg, and can all be related to one another through a cellular family tree. In simpler organisms, such as the worm C. elegans, this cellular family tree has been fully mapped, but the cellular family tree of a human is many times larger and more complex.

In the past, MIT professor and Whitehead Institute for Biomedical Research member Jonathan Weissman and other researchers developed lineage tracing methods to track and reconstruct the family trees of cell divisions in model organisms in order to understand more about the relationships between cells and how they assemble into tissues, organs, and — in some cases — tumors. These methods could help to answer many questions about how organisms develop and diseases like cancer are initiated and progress.

Now, Weissman and colleagues have developed an advanced lineage tracing tool that not only captures an accurate family tree of cell divisions, but also combines that with spatial information: identifying where each cell ends up within a tissue. The researchers used their tool, PEtracer, to observe the growth of metastatic tumors in mice. Combining lineage tracing and spatial data provided the researchers with a detailed view of how elements intrinsic to the cancer cells and from their environments influenced tumor growth, as Weissman and postdocs in his lab Luke Koblan, Kathryn Yost, and Pu Zheng, and graduate student William Colgan share in a paper published in the journal Science on July 24.

“Developing this tool required combining diverse skill sets through the sort of ambitious interdisciplinary collaboration that’s only possible at a place like Whitehead Institute,” says Weissman, who is also a Howard Hughes Medical Institute investigator. “Luke came in with an expertise in genetic engineering, Pu in imaging, Katie in cancer biology, and William in computation, but the real key to their success was their ability to work together to build PEtracer.”

“Understanding how cells move in time and space is an important way to look at biology, and here we were able to see both of those things in high resolution. The idea is that by understanding both a cell’s past and where it ends up, you can see how different factors throughout its life influenced its behaviors. In this study, we use these approaches to look at tumor growth, though in principle we can now begin to apply these tools to study other biology of interest, like embryonic development,” Koblan says.

Designing a tool to track cells in space and time

PEtracer tracks cells’ lineages by repeatedly adding short, predetermined codes to the DNA of cells over time. Each piece of code, called a lineage tracing mark, is made up of five bases, the building blocks of DNA. These marks are inserted using a gene editing technology called prime editing, which directly rewrites stretches of DNA with minimal undesired byproducts. Over time, each cell acquires more lineage tracing marks, while also maintaining the marks of its ancestors. The researchers can then compare cells’ combinations of marks to figure out relationships and reconstruct the family tree.

“We used computational modeling to design the tool from first principles, to make sure that it was highly accurate, and compatible with imaging technology. We ran many simulations to land on the optimal parameters for a new lineage tracing tool, and then engineered our system to fit those parameters,” Colgan says.

When the tissue — in this case, a tumor growing in the lung of a mouse — had sufficiently grown, the researchers collected these tissues and used advanced imaging approaches to look at each cell’s lineage relationship to other cells via the lineage tracing marks, along with its spatial position within the imaged tissue and its identity (as determined by the levels of different RNAs expressed in each cell). PEtracer is compatible with both imaging approaches and sequencing methods that capture genetic information from single cells.

“Making it possible to collect and analyze all of this data from the imaging was a large challenge,” Zheng says. “What’s particularly exciting to me is not just that we were able to collect terabytes of data, but that we designed the project to collect data that we knew we could use to answer important questions and drive biological discovery.”

Reconstructing the history of a tumor

Combining the lineage tracing, gene expression, and spatial data let the researchers understand how the tumor grew. They could tell how closely related neighboring cells are and compare their traits. Using this approach, the researchers found that the tumors they were analyzing were made up of four distinct modules, or neighborhoods, of cells.

The tumor cells closest to the lung, the most nutrient-dense region, were the most fit, meaning their lineage history indicated the highest rate of cell division over time. Fitness in cancer cells tends to correlate to how aggressively tumors will grow.

The cells at the “leading edge” of the tumor, the far side from the lung, were more diverse and not as fit. Below the leading edge was a low-oxygen neighborhood of cells that might once have been leading edge cells, now trapped in a less-desirable spot. Between these cells and the lung-adjacent cells was the tumor core, a region with both living and dead cells, as well as cellular debris.

The researchers found that cancer cells across the family tree were equally likely to end up in most of the regions, with the exception of the lung-adjacent region, where a few branches of the family tree dominated. This suggests that the cancer cells’ differing traits were heavily influenced by their environments, or the conditions in their local neighborhoods, rather than their family history. Further evidence of this point was that expression of certain fitness-related genes, such as Fgf1/Fgfbp1, correlated to a cell’s location, rather than its ancestry. However, lung-adjacent cells also had inherited traits that gave them an edge, including expression of the fitness-related gene Cldn4­ — showing that family history influenced outcomes as well.

These findings demonstrate how cancer growth is influenced both by factors intrinsic to certain lineages of cancer cells and by environmental factors that shape the behavior of cancer cells exposed to them.

“By looking at so many dimensions of the tumor in concert, we could gain insights that would not have been possible with a more limited view,” Yost says. “Being able to characterize different populations of cells within a tumor will enable researchers to develop therapies that target the most aggressive populations more effectively.”

“Now that we’ve done the hard work of designing the tool, we’re excited to apply it to look at all sorts of questions in health and disease, in embryonic development, and across other model species, with an eye toward understanding important problems in human health,” Koblan says. “The data we collect will also be useful for training AI models of cellular behavior. We’re excited to share this technology with other researchers and see what we all can discover.”

Salt creeping, a phenomenon that occurs in both natural and industrial processes, describes the collection and migration of salt crystals from evaporating solutions onto surfaces. Once they start collecting, the crystals climb, spreading away from the solution. This creeping behavior, according to researchers, can cause damage or be harnessed for good, depending on the context. New research published June 30 in the journal Langmuir is the first to show salt creeping at a single-crystal scale and beneath a liquid’s meniscus.

“The work not only explains how salt creeping begins, but why it begins and when it does,” says Joseph Phelim Mooney, a postdoc in the MIT Device Research Laboratory and one of the authors of the new study. “We hope this level of insight helps others, whether they’re tackling water scarcity, preserving ancient murals, or designing longer-lasting infrastructure.”

The work is the first to directly visualize how salt crystals grow and interact with surfaces underneath a liquid meniscus, something that’s been theorized for decades but never actually imaged or confirmed at this level, and it offers fundamental insights that could impact a wide range of fields — from mineral extraction and desalination to anti-fouling coatings, membrane design for separation science, and even art conservation, where salt damage is a major threat to heritage materials.

In civil engineering applications, for example, the research can help explain why and when salt crystals start growing across surfaces like concrete, stone, or building materials. “These crystals can exert pressure and cause cracking or flaking, reducing the long-term durability of structures,” says Mooney. “By pinpointing the moment when salt begins to creep, engineers can better design protective coatings or drainage systems to prevent this form of degradation.”

For a field like art conservation, where salt can be devastating to murals, frescoes, and ancient artifacts, often forming beneath the surface before visible damage appears, the work can help identify the exact conditions that cause salt to start moving and spreading, allowing conservators to act earlier and more precisely to protect heritage objects.

The work began during Mooney’s Marie Curie Fellowship at MIT. “I was focused on improving desalination systems and quickly ran into [salt buildup as] a major roadblock,” he says. “[Salt] was everywhere, coating surfaces, clogging flow paths, and undermining the efficiency of our designs. I realized we didn’t fully understand how or why salt starts creeping across surfaces in the first place.”

That experience led Mooney to team up with colleagues to dig into the fundamentals of salt crystallization at the air–liquid–solid interface. “We wanted to zoom in, to really see the moment salt begins to move, so we turned to in situ X-ray microscopy,” he says. “What we found gave us a whole new way to think about surface fouling, material degradation, and controlled crystallization.”

The new research may, in fact, allow better control of a crystallization processes required to remove salt from water in zero-liquid discharge systems. It can also be used to explain how and when scaling happens on equipment surfaces, and may support emerging climate technologies that depend on smart control of evaporation and crystallization.

The work also supports mineral and salt extraction applications, where salt creeping can be both a bottleneck and an opportunity. In these applications, Mooney says, “by understanding the precise physics of salt formation at surfaces, operators can optimize crystal growth, improving recovery rates and reducing material losses.”

Mooney’s co-authors on the paper include fellow MIT Device Lab researchers Omer Refet Caylan, Bachir El Fil (now an associate professor at Georgia Tech), and Lenan Zhang (now an associate professor at Cornell University); Jeff Punch and Vanessa Egan of the University of Limerick; and Jintong Gao of Cornell.

The research was conducted using in situ X-ray microscopy. Mooney says the team’s big realization moment occurred when they were able to observe a single salt crystal pinning itself to the surface, which kicked off a cascading chain reaction of growth.

“People had speculated about this, but we captured it on X-ray for the first time. It felt like watching the microscopic moment where everything tips, the ignition points of a self-propagating process,” says Mooney. “Even more surprising was what followed: The salt crystal didn’t just grow passively to fill the available space. It pierced through the liquid-air interface and reshaped the meniscus itself, setting up the perfect conditions for the next crystal. That subtle, recursive mechanism had never been visually documented before — and seeing it play out in real time completely changed how we thought about salt crystallization.”

The paper, “In Situ X-ray Microscopy Unraveling the Onset of Salt Creeping at a Single-Crystal Level,” is available now in the journal Langmuir. Research was conducted in MIT.nano. 

If you rotate an image of a molecular structure, a human can tell the rotated image is still the same molecule, but a machine-learning model might think it is a new data point. In computer science parlance, the molecule is “symmetric,” meaning the fundamental structure of that molecule remains the same if it undergoes certain transformations, like rotation.

If a drug discovery model doesn’t understand symmetry, it could make inaccurate predictions about molecular properties. But despite some empirical successes, it’s been unclear whether there is a computationally efficient method to train a good model that is guaranteed to respect symmetry.

A new study by MIT researchers answers this question, and shows the first method for machine learning with symmetry that is provably efficient in terms of both the amount of computation and data needed.

These results clarify a foundational question, and they could aid researchers in the development of more powerful machine-learning models that are designed to handle symmetry. Such models would be useful in a variety of applications, from discovering new materials to identifying astronomical anomalies to unraveling complex climate patterns.

“These symmetries are important because they are some sort of information that nature is telling us about the data, and we should take it into account in our machine-learning models. We’ve now shown that it is possible to do machine-learning with symmetric data in an efficient way,” says Behrooz Tahmasebi, an MIT graduate student and co-lead author of this study.

He is joined on the paper by co-lead author and MIT graduate student Ashkan Soleymani; Stefanie Jegelka, an associate professor of electrical engineering and computer science (EECS) and a member of the Institute for Data, Systems, and Society (IDSS) and the Computer Science and Artificial Intelligence Laboratory (CSAIL); and senior author Patrick Jaillet, the Dugald C. Jackson Professor of Electrical Engineering and Computer Science and a principal investigator in the Laboratory for Information and Decision Systems (LIDS). The research was recently presented at the International Conference on Machine Learning.

Studying symmetry

Symmetric data appear in many domains, especially the natural sciences and physics. A model that recognizes symmetries is able to identify an object, like a car, no matter where that object is placed in an image, for example.

Unless a machine-learning model is designed to handle symmetry, it could be less accurate and prone to failure when faced with new symmetric data in real-world situations. On the flip side, models that take advantage of symmetry could be faster and require fewer data for training.

But training a model to process symmetric data is no easy task.

One common approach is called data augmentation, where researchers transform each symmetric data point into multiple data points to help the model generalize better to new data. For instance, one could rotate a molecular structure many times to produce new training data, but if researchers want the model to be guaranteed to respect symmetry, this can be computationally prohibitive.

An alternative approach is to encode symmetry into the model’s architecture. A well-known example of this is a graph neural network (GNN), which inherently handles symmetric data because of how it is designed.

“Graph neural networks are fast and efficient, and they take care of symmetry quite well, but nobody really knows what these models are learning or why they work. Understanding GNNs is a main motivation of our work, so we started with a theoretical evaluation of what happens when data are symmetric,” Tahmasebi says.

They explored the statistical-computational tradeoff in machine learning with symmetric data. This tradeoff means methods that require fewer data can be more computationally expensive, so researchers need to find the right balance.

Building on this theoretical evaluation, the researchers designed an efficient algorithm for machine learning with symmetric data.

Mathematical combinations

To do this, they borrowed ideas from algebra to shrink and simplify the problem. Then, they reformulated the problem using ideas from geometry that effectively capture symmetry.

Finally, they combined the algebra and the geometry into an optimization problem that can be solved efficiently, resulting in their new algorithm.

“Most of the theory and applications were focusing on either algebra or geometry. Here we just combined them,” Tahmasebi says.

The algorithm requires fewer data samples for training than classical approaches, which would improve a model’s accuracy and ability to adapt to new applications.

By proving that scientists can develop efficient algorithms for machine learning with symmetry, and demonstrating how it can be done, these results could lead to the development of new neural network architectures that could be more accurate and less resource-intensive than current models.

Scientists could also use this analysis as a starting point to examine the inner workings of GNNs, and how their operations differ from the algorithm the MIT researchers developed.

“Once we know that better, we can design more interpretable, more robust, and more efficient neural network architectures,” adds Soleymani.

This research is funded, in part, by the National Research Foundation of Singapore, DSO National Laboratories of Singapore, the U.S. Office of Naval Research, the U.S. National Science Foundation, and an Alexander von Humboldt Professorship.

Music technology took center stage at MIT during “FUTURE PHASES,” an evening of works for string orchestra and electronics, presented by the MIT Music Technology and Computation Graduate Program as part of the 2025 International Computer Music Conference (ICMC). 

The well-attended event was held last month in the Thomas Tull Concert Hall within the new Edward and Joyce Linde Music Building. Produced in collaboration with the MIT Media Lab’s Opera of the Future Group and Boston’s self-conducted chamber orchestra A Far Cry, “FUTURE PHASES” was the first event to be presented by the MIT Music Technology and Computation Graduate Program in MIT Music’s new space.

“FUTURE PHASES” offerings included two new works by MIT composers: the world premiere of “EV6,” by MIT Music’s Kenan Sahin Distinguished Professor Evan Ziporyn and professor of the practice Eran Egozy; and the U.S. premiere of “FLOW Symphony,” by the MIT Media Lab’s Muriel R. Cooper Professor of Music and Media Tod Machover. Three additional works were selected by a jury from an open call for works: “The Wind Will Carry Us Away,” by Ali Balighi; “A Blank Page,” by Celeste Betancur Gutiérrez and Luna Valentin; and “Coastal Portrait: Cycles and Thresholds,” by Peter Lane. Each work was performed by Boston’s own multi-Grammy-nominated string orchestra, A Far Cry.

“The ICMC is all about presenting the latest research, compositions, and performances in electronic music,” says Egozy, director of the new Music Technology and Computation Graduate Program at MIT. When approached to be a part of this year’s conference, “it seemed the perfect opportunity to showcase MIT’s commitment to music technology, and in particular the exciting new areas being developed right now: a new master’s program in music technology and computation, the new Edward and Joyce Linde Music Building with its enhanced music technology facilities, and new faculty arriving at MIT with joint appointments between MIT Music and Theater Arts (MTA) and the Department of Electrical Engineering and Computer Science (EECS).” These recently hired professors include Anna Huang, a keynote speaker for the conference and creator of the machine learning model Coconet that powered Google’s first AI Doodle, the Bach Doodle.

Egozy emphasizes the uniqueness of this occasion: “You have to understand that this is a very special situation. Having a full 18-member string orchestra [A Far Cry] perform new works that include electronics does not happen very often. In most cases, ICMC performances consist either entirely of electronics and computer-generated music, or perhaps a small ensemble of two-to-four musicians. So the opportunity we could present to the larger community of music technology was particularly exciting.”

To take advantage of this exciting opportunity, an open call was put out internationally to select the other pieces that would accompany Ziporyn and Egozy’s “EV6” and Machover’s “FLOW Symphony.” Three pieces were selected from a total of 46 entries to be a part of the evening’s program by a panel of judges that included Egozy, Machover, and other distinguished composers and technologists.

“We received a huge variety of works from this call,” says Egozy. “We saw all kinds of musical styles and ways that electronics would be used. No two pieces were very similar to each other, and I think because of that, our audience got a sense of how varied and interesting a concert can be for this format. A Far Cry was really the unifying presence. They played all pieces with great passion and nuance. They have a way of really drawing audiences into the music. And, of course, with the Thomas Tull Concert Hall being in the round, the audience felt even more connected to the music.”

Egozy continues, “we took advantage of the technology built into the Thomas Tull Concert Hall, which has 24 built-in speakers for surround sound allowing us to broadcast unique, amplified sound to every seat in the house. Chances are that every person might have experienced the sound slightly differently, but there was always some sense of a multidimensional evolution of sound and music as the pieces unfolded.”

The five works of the evening employed a range of technological components that included playing synthesized, prerecorded, or electronically manipulated sounds; attaching microphones to instruments for use in real-time signal processing algorithms; broadcasting custom-generated musical notation to the musicians; utilizing generative AI to process live sound and play it back in interesting and unpredictable ways; and audience participation, where spectators use their cellphones as musical instruments to become a part of the ensemble.

Ziporyn and Egozy’s piece, “EV6,” took particular advantage of this last innovation: “Evan and I had previously collaborated on a system called Tutti, which means ‘together’ in Italian. Tutti gives an audience the ability to use their smartphones as musical instruments so that we can all play together.” Egozy developed the technology, which was first used in the MIT Campaign for a Better World in 2017. The original application involved a three-minute piece for cellphones only. “But for this concert,” Egozy explains, “Evan had the idea that we could use the same technology to write a new piece — this time, for audience phones and a live string orchestra as well.”

To explain the piece’s title, Ziporyn says, “I drive an EV6; it’s my first electric car, and when I first got it, it felt like I was driving an iPhone. But of course it’s still just a car: it’s got wheels and an engine, and it gets me from one place to another. It seemed like a good metaphor for this piece, in which a lot of the sound is literally played on cellphones, but still has to work like any other piece of music. It’s also a bit of an homage to David Bowie’s song ‘TVC 15,’ which is about falling in love with a robot.”

Egozy adds, “We wanted audience members to feel what it is like to play together in an orchestra. Through this technology, each audience member becomes a part of an orchestral section (winds, brass, strings, etc.). As they play together, they can hear their whole section playing similar music while also hearing other sections in different parts of the hall play different music. This allows an audience to feel a responsibility to their section, hear how music can move between different sections of an orchestra, and experience the thrill of live performance. In ‘EV6,’ this experience was even more electrifying because everyone in the audience got to play with a live string orchestra — perhaps for the first time in recorded history.”

After the concert, guests were treated to six music technology demonstrations that showcased the research of undergraduate and graduate students from both the MIT Music program and the MIT Media Lab. These included a gamified interface for harnessing just intonation systems (Antonis Christou); insights from a human-AI co-created concert (Lancelot Blanchard and Perry Naseck); a system for analyzing piano playing data across campus (Ayyub Abdulrezak ’24, MEng ’25); capturing music features from audio using latent frequency-masked autoencoders (Mason Wang); a device that turns any surface into a drum machine (Matthew Caren ’25); and a play-along interface for learning traditional Senegalese rhythms (Mariano Salcedo ’25). This last example led to the creation of Senegroove, a drumming-based application specifically designed for an upcoming edX online course taught by ethnomusicologist and MIT associate professor in music Patricia Tang, and world-renowned Senegalese drummer and MIT lecturer in music Lamine Touré, who provided performance videos of the foundational rhythms used in the system.

Ultimately, Egozy muses, “'FUTURE PHASES' showed how having the right space — in this case, the new Edward and Joyce Linde Music Building — really can be a driving force for new ways of thinking, new projects, and new ways of collaborating. My hope is that everyone in the MIT community, the Boston area, and beyond soon discovers what a truly amazing place and space we have built, and are still building here, for music and music technology at MIT.”