SPARKLE: evidence-constrained correction of local RNA leakage in high-resolution spatial transcriptomics

To understand the landscape of life inside a tissue, scientists have developed a way to read the genetic instructions of thousands of individual cells at once, while keeping track of exactly where each cell sits in the body. This field, known as spatial transcriptomics, allows researchers to map which genes are active in specific neighborhoods, revealing how different cell types organize themselves to build organs or fight disease. However, a persistent problem has clouded these maps. When scientists prepare tissue samples for this kind of analysis, they must soften the tissue to let the genetic material out. In doing so, some of the genetic messages, or RNA, escape their original homes and drift into the spaces between cells. This leakage creates a fog of false signals, making it look as though a cell is producing genetic instructions it never actually made, which can lead to confusion about what a cell really is and what it is doing with its neighbors.

A new method called SPARKLE offers a way to clear this fog without needing a reference guide from outside the sample. The researchers behind this work realized that the very act of leakage leaves a trace. When RNA escapes a cell, it does not vanish; it lands on the empty spaces surrounding the cell, the areas that do not belong to any specific tissue structure. By looking at these empty spaces, the method can measure how far the genetic material tends to travel and how much of it leaks out. Instead of guessing or relying on external data, SPARKLE uses the evidence found right there in the sample to calculate a correction. It builds a model of how the RNA spreads, using the signals found in the empty zones to estimate how much of the signal inside a cell is actually its own and how much is just borrowed from a neighbor.

The approach works by fitting a mathematical pattern to the observations found outside the cell boundaries. This pattern helps the software determine a scale for how far RNA travels in that specific tissue sample and how much of it leaks for each specific gene. Crucially, the method only corrects the genes where the evidence from the empty spaces is strong enough to support the calculation. For genes that are highly active in their source cells, the method applies a careful, conservative adjustment to ensure it does not accidentally erase the true signal. This allows the software to distinguish between a cell that is genuinely producing a protein and a cell that is merely reflecting the presence of a neighbor.

The researchers tested this idea across ten different simulated scenarios designed to mimic the messy reality of biological tissue. In nine out of ten of these tests, the method produced the most accurate maps of cell activity and the smallest errors compared to the known truth. To see if it worked in the real world, the team applied SPARKLE to samples from an axolotl brain, a mouse brain, and human ovarian cancer. In each case, the method successfully removed the false signals that had appeared in neighboring cells, while keeping the true signals intact in the cells where the genes originated. This correction revealed a clearer picture of the tissue, bringing the results into closer agreement with independent data from single-cell studies. In the human ovarian cancer samples, the method uncovered a specific type of communication between cells that had been hidden by the diffusion of collagen, showing that fibroblast cells were sending signals that were previously obscured.

The tool proved to be both fast and reliable. The time it took to process the data grew in a steady, predictable way as the size of the tissue area increased, and the process could be sped up even further using specialized computer hardware. The results remained consistent even when the researchers adjusted the size of the areas they analyzed or the scale of the spread they assumed. By using the evidence contained within each sample to fix the problem of leakage, SPARKLE provides a reference-free way to sharpen the view of cellular life. This improvement strengthens the ability to identify where different cell types live, how they are grouped into tissue compartments, and how they communicate with one another, offering a clearer foundation for understanding the complex architecture of the body.