Duke-NUS scientists build new virtual tissue tools to map how cells talk in disease

Scientists at Duke-NUS Medical School have developed two powerful computational tools that could transform how researchers study the “conversations” between cells inside the body. The tools, called sCCIgen and QuadST, help scientists understand both where cells are located in tissues and how they communicate through genetic activity and chemical signals.

Each study is published in leading peer-reviewed journals:

• sCCIgen, described in Genome Biology (Springer Nature), introduces the first simulator capable of generating realistic, multi-layered virtual tissues that fully capture cell locations, gene activity patterns, and communication networks.
• QuadST, detailed in Genome Research (Cold Spring Harbor Laboratory Press), showcases the tool’s ability to detect cell-to-cell communication signals directly from spatial transcriptomics data, revealing genes that change as cells interact in healthy and diseased tissues.

Cells constantly send and receive signals that help the body grow, stay healthy, and defend against threats. When these signals fail, disease often follows. For example, disrupted communication between nerve cells can lead to memory loss and cognitive decline, as seen in Alzheimer’s disease. Cancer cells can also exploit signalling pathways—sending misleading cues to immune cells that allow tumours to evade detection. Understanding which cells communicate and what messages they exchange, is therefore a central goal in biomedical research.

Despite its importance, cellular communication has been notoriously difficult to study. Traditional experimental approaches struggle to capture interactions between cells within their native tissue environment. Spatial transcriptomics has begun to change this landscape. By showing not only which genes are active in each cell but also exactly where those cells sit within a tissue, it enables scientists to create detailed maps of cellular organisation and gene activity in the tissue. These maps are rich in information, but so complex that existing computational tools struggle to analyse them reliably.

A key obstacle is the lack of “ground truth” in real biological data. Researchers rarely know which cells are truly interacting, making it difficult to evaluate whether computational methods are correctly identifying communication signals.

The first new tool, called sCCIgen, addresses this gap by generating realistic, computer-simulated tissues. These virtual models specify every cell’s position, gene activity profile, and communication links. Because the underlying biology in these simulations is fully known, scientists can test whether their methods for detecting cell communication are accurate. In effect, sCCIgen functions like a controlled “practice lab”, allowing researchers to refine and validate their computational strategies.

Associate Professor Xiaoyu Song of Duke-NUS’ Centre for Quantitative Medicine, lead author of both studies, said: “Just as flight simulators give pilots a safe place to practice, sCCIgen gives scientists a controlled environment to test whether their computational tools can really detect how cells ‘talk’ to each other. This will help accelerate discoveries in cancer, immunology, and neuroscience.”

Unlike earlier simulators that referenced limited data, sCCIgen can incorporate diverse types of information, from single-cell maps to datasets gathered from other technologies. It can model different interaction patterns, such as clustered cell neighbourhoods, genes that change activity depending on how far apart cells are, and gene pairs that activate when neighbouring cells interact. This flexibility enables researchers design more realistic experiments and evaluate how well their analytical tools capture genuine communication signals. For example, sCCIgen can be used to explore which immune cells recognise a tumour and which are being suppressed, informing experiments and therapies.

While sCCIgen generates virtual tissues, the second tool, QuadST, analyses real tissue data to reveal which genes and cell types are communicating. Earlier methods often missed key interactions because they relied on estimating how cells interact. QuadST improves on this by modelling how gene activity changes progressively with the distance between two cell types. It is also designed to handle noisy or incomplete datasets, a common challenge in spatial transcriptomics.

When the researchers applied QuadST to brain tissue, they identified hundreds of genes whose activity shifts when specific nerve cells interact. Many of these genes are associated with synapses, the junctions through which nerve cells exchange signals. These insights could help scientists understand how neuronal communication is altered in conditions like epilepsy or neurodegenerative diseases. QuadST can also be used to study tumours, clarifying how cancer cells interact with immune cells, and why some patients respond better to treatment than others. By pinpointing genes that change when communication patterns differ in disease, QuadST helps uncover mechanisms that were previously difficult to detect.

Assoc Prof Song explained: “Cells live in busy neighbourhoods. They don’t send signals at random but respond to what their neighbours are doing. QuadST helps us trace these interactions accurately. It shows which genes are changing because of cell communication instead of by chance, helping researchers uncover hidden disease mechanisms that were too difficult to detect before.”

Professor Patrick Tan, Senior Vice Dean for Research at Duke-NUS, said: “Together, sCCIgen and QuadST give scientists a reliable process for studying cell communication. Researchers can first create virtual tissues with sCCIgen to test and refine their methods, then use QuadST to examine real tissue data and pinpoint exactly which genes are involved in cell interactions. Using the two tools in this way helps ensure the results are accurate and makes it easier to uncover new insights about how cells work in health and disease.”

The team plans to expand sCCIgen to simulate proteins and other molecular interactions, and to use QuadST to build a reference database of genes involved in cell-cell communication. This resource will help scientists compare findings across tissues and diseases, accelerate discovery, and support the development of more precise treatments.

Duke-NUS is a biomedical research powerhouse, combining basic scientific research with translational expertise to deepen understanding of common diseases and develop new treatment approaches.