Phylospatial optimal transport for single-cell and spatial transcriptomics data integration
The paper introduces SpOTr, an optimal transport-based method that integrates single-cell and spatial transcriptomics data by leveraging evolutionary tree concordance to accurately map cancer clone lineages to their spatial locations within tumors.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Tumors are not uniform lumps of identical cells; they are complex, evolving ecosystems where different groups of cancer cells, known as clones, compete and cooperate. These clones arise as cells divide and accumulate genetic changes, creating a family tree of distinct lineages within a single patient. Understanding how these lineages spread and interact with their surroundings is crucial, because the physical location of a specific cancer clone often dictates how it behaves and whether it will survive treatment. However, scientists have long faced a difficult choice when studying these tumors. They can either look at the tumor as a whole, preserving the map of where every cell sits but losing the ability to see individual cell details, or they can break the tumor apart to study individual cells in high definition, only to lose the map entirely. This trade-off has made it incredibly hard to connect the genetic history of a cancer clone to its actual position inside the body.
A new approach called SpOTr offers a way to bridge this gap by combining two different types of data into a single, coherent picture. The method takes high-resolution genetic information from cells that have been removed from the tumor and matches it to a lower-resolution map of the tumor tissue that still shows where cells are located. The key insight driving this work is that closely related cancer cells tend to stay near each other as they grow. By using this rule of proximity, the researchers developed a mathematical framework that can place individual cancer cells back onto the tissue map, even when the map itself is blurry and contains mixtures of many different cell types. This allows scientists to see not just where cancer cells are, but how the family tree of the tumor unfolds across the physical space of the organ.
The researchers tested this idea by creating simulated tumors where they knew the exact location of every cell and every clone. In these computer-generated scenarios, the clones were designed to be very similar to one another, making them nearly impossible to tell apart just by looking at their genetic activity. When the team applied SpOTr to these simulations, it successfully reconstructed the territories of the different clones, whereas other existing methods that relied only on genetic activity failed to distinguish between them. The method proved particularly powerful when the clones were hard to differentiate, using the known family relationships to guide the cells to their correct neighborhoods. In one test, the method correctly identified the spatial boundaries of three major clones and parts of two smaller ones, while other tools assigned most of the tissue to just the largest group, missing the others entirely.
To see if this worked with real biological data, the team applied SpOTr to a tumor where the lineage of every cell had been tracked using a special barcode system. In this case, they had a ground truth: they knew exactly which part of the tumor each cell belonged to. When they compared the results of SpOTr against the known reality, the method accurately recovered the main regions occupied by different branches of the cancer family tree. It kept the upper and lower parts of the tumor distinct, matching the actual layout, while other methods either mixed the regions together or assigned the entire lower section to a single group. This confirmed that the approach could handle the complexity of real tissue, where cells are mixed together in small clusters and the genetic signals are often too weak to be read clearly on their own.
The researchers then applied the method to a human colorectal cancer that had spread to the liver, a scenario where no pre-existing map of the cell lineages existed. They identified two main cancer groups and two smaller sub-groups within one of them. Using the family tree derived from genetic mutations, SpOTr was able to separate these sub-groups and place them in different parts of the tissue. One sub-group was found in the nests of cells on the left side of the sample, while the other was located in a band across the top. This level of detail was not possible with standard methods, which treated the entire group as a single block. By placing the cells back into their spatial context, the researchers could draw the family tree directly onto the image of the tissue, showing how the cancer evolved from a common ancestor near one side of the tumor to distinct, separated regions on the other.
This work suggests that the history of a tumor is written in its geography. By combining the high-definition genetic data of individual cells with the spatial context of the tissue, SpOTr allows scientists to see the evolutionary journey of cancer as it happens in the body. The method does not require the tissue to be broken apart to be studied, nor does it require the cells to be perfectly distinct genetically. Instead, it uses the simple biological rule that relatives tend to stay close together to solve a complex puzzle. In doing so, it provides a new way to visualize the hidden architecture of tumors, offering a clearer view of how cancer grows, spreads, and adapts to its environment.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.