SpaceBlocks: an expert-supervised workflow for cohort-level spatial transcriptomics analysis
SpaceBlocks is a reproducible, expert-supervised Snakemake workflow that homogenizes heterogeneous single-cell resolution spatial transcriptomics data across multiple platforms to enable scalable cohort-level analysis, as demonstrated by its ability to identify novel cell types, marker expression patterns, and disease-associated transcriptional differences in diverse biological datasets.
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 This is an AI-generated explanation of a preprint that has not been peer-reviewed.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
Tissue is not a uniform soup of cells; it is a structured city where location defines function. For decades, biologists could read the genetic instructions inside a cell, but they had to grind the tissue into a paste to do it, losing the map of where each cell lived. A newer technology, called spatial transcriptomics, solves this by reading the genetic code while keeping the cell in its original neighborhood. This allows scientists to see how cells talk to their neighbors and organize themselves into functional districts, much like understanding a city by seeing which shops, schools, and homes sit next to one another. However, this new way of looking at tissue comes with a complication: different machines that take these pictures speak different languages. One machine might output its data in a format that another machine cannot read, and the software tools used to analyze them often work in isolation. When researchers try to study a large group of patients together—a cohort—they face a chaotic mix of file types and manual steps that are hard to repeat or verify.
To bring order to this complexity, a team of researchers has built a new digital workflow called SpaceBlocks. Think of this workflow as a universal translator and traffic controller for tissue data. Instead of letting every researcher build their own custom tools for every new experiment, SpaceBlocks provides a single, standardized path that any computer can follow. It takes the raw, messy output from different high-tech microscopes and converts it into a clean, common format that all downstream analysis tools can understand. Crucially, this system does not try to automate every decision away. The researchers designed it to pause and ask for human expertise at key moments, such as deciding which cells are healthy enough to keep or how to label a group of cells based on their genetic profile. This ensures that the final results are not just the output of a black-box computer program, but a verified product of scientific judgment.
The team tested this system on a large collection of tissue samples from three different types of advanced microscopes, covering a total of 37 samples from public databases. These samples included tissue from patients with multiple sclerosis and from patients with colorectal cancer. The workflow successfully processed all of them, proving it could handle the different file formats and the varying sizes of the datasets without breaking. In one instance involving multiple sclerosis, the system helped researchers identify specific genetic patterns in different parts of the brain lesions that had been missed in previous analyses. It revealed that certain genes were active in specific neighborhoods of the tissue, providing a clearer picture of how the disease affects different areas. In another test using colorectal cancer samples, the workflow helped researchers spot three distinct types of cells that a previous, widely used analysis had overlooked. These included immune cells and nerve-like cells that were hiding in plain sight within the tumor tissue.
Beyond just finding new cells, the system showed how to compare healthy tissue against diseased tissue in a way that accounts for the unique structure of each sample. By grouping cells into "pseudobulk" profiles—essentially combining the genetic signals of many cells in a specific area to get a stronger, clearer signal—the researchers could pinpoint exactly which genes were turning on or off in the disease state. The system ran these complex comparisons automatically once the human experts had set the rules, generating detailed maps and charts that showed the differences between healthy and malignant tissues. The results were consistent across the different machines, suggesting that the workflow is robust enough to be used by other scientists who want to study large groups of patients without getting lost in technical details.
The researchers emphasize that this tool is not a replacement for human insight but a framework to support it. By making the steps of the analysis transparent and repeatable, SpaceBlocks allows scientists to trust that their findings are real and not just an artifact of a specific software setting. The system is freely available to anyone, and the code is open for others to improve or adapt for new types of microscopes. In the future, the team plans to add support for even more imaging technologies and to make the initial steps of the process fully open-source, removing any reliance on proprietary software. For now, SpaceBlocks stands as a practical solution to a growing problem in biology: how to make sense of the vast, complex, and varied data coming from the new generation of tissue-mapping machines, ensuring that the story of the tissue is told clearly and accurately.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.