Biological tissues are complex environments composed of many different cell types. When scientists study a tissue sample, the genetic information obtained is a mixture of data from all these individual cells, creating a challenge in understanding the role each one plays. To address this, researchers use computational methods to digitally separate these mixed signals.
One such method is xCell deconvolution, a computational tool that analyzes gene expression data from a tissue sample to estimate the abundance of various cell types. This provides researchers with a detailed portrait of a tissue’s cellular composition from a single biological sample. The resulting insights are valuable for understanding complex biological processes and diseases.
The Science of Digital Cell Sorting
When a biological tissue sample is analyzed for its genetic activity, the resulting data represents a “bulk” measurement. This data is an average of the gene expression from every cell contained within that sample. Dissecting this composite signal is a challenge because different cells have different functions and patterns of gene activity.
To overcome this, scientists rely on gene expression signatures, which are unique patterns of gene activity that function like a fingerprint for a specific cell. For instance, an immune T-cell will have a different set of active genes compared to a skin cell or a neuron. By identifying these unique signatures within the bulk data, it’s possible to infer which cell types are present and in what relative amounts.
This process is analogous to listening to an orchestra and identifying which instruments are playing. Just as a trained ear can pick out a violin from a trumpet based on their unique sounds, computational tools can distinguish a B-cell from a macrophage based on their distinct gene expression patterns. This “digital cell sorting” allows for a detailed understanding of the cellular landscape of a tissue without physically separating the cells.
This approach can be applied to data from high-throughput methods like RNA-sequencing, which measure thousands of genes at once. By comparing the bulk gene expression profile against a reference library of known cell-type-specific signatures, researchers can deconstruct the mixed signal. This provides a quantitative estimation of the cellular makeup of the original tissue.
The xCell Algorithm and Its Signatures
The xCell tool is distinguished by its use of a vast reference library containing gene signatures from pure human cell types. This database was built by curating thousands of publicly available gene expression profiles from isolated cell populations. The result is a collection of 489 reliable gene signatures used to identify 64 different immune, stromal, and progenitor cell types.
The process begins when a user provides bulk gene expression data from a tissue sample. The xCell algorithm then employs single-sample Gene Set Enrichment Analysis (ssGSEA). This analysis calculates a raw enrichment score for each of the 64 cell types by assessing how strongly its gene signatures are present in the user’s bulk data.
A unique feature of the xCell method is its spillover compensation technique. Different cell types, especially closely related ones like T-cell subtypes, often share genes in their expression signatures. This overlap can lead to inaccuracies, where the signal from one cell type artificially inflates the score of another.
To correct for this, xCell uses data from simulated tissue mixtures to learn the dependencies between different cell types. It then applies a compensation algorithm to adjust the enrichment scores. This process reduces the influence of shared genes, isolating the unique signals for each cell type and improving the accuracy of the final estimate.
The latest version, xCell 2.0, further refines this by integrating cell ontology to better handle hierarchical relationships between cell types. This prevents incorrect adjustments between a parent cell type like “T-cells” and a child cell type like “CD4+ T-cells.”
Applications in Biomedical Research
xCell is a widely used tool in biomedical research, particularly in oncology and immunology. One of its primary applications is in the analysis of the tumor microenvironment (TME). The TME is a complex ecosystem of cancer cells, blood vessels, structural cells, and immune cells that influences tumor growth and response to treatment.
Researchers use xCell to analyze bulk gene expression data from tumor biopsies to estimate the abundance of different immune cells infiltrating a tumor. For instance, determining the levels of cytotoxic T-cells versus regulatory T-cells provides insight into the anti-tumor immune response. This information is useful for predicting how a patient might respond to immunotherapies, such as immune checkpoint inhibitors.
Beyond cancer, xCell is applied to study cellular dynamics in a variety of other diseases. In autoimmune disorders, researchers can use it to characterize the types of immune cells present in inflamed tissues to understand the disease mechanisms. It can also be used to track changes in immune cell populations during an infection or in response to a vaccine.
The tool’s utility extends to preclinical research as well. Scientists analyzing patient-derived xenograft (PDX) models, where human tumors are grown in mice, use xCell to characterize the TME of these models. This allows them to select more clinically relevant models for testing new therapies.
Interpreting and Validating xCell Scores
When working with xCell data, it is important to understand the results. The tool produces “enrichment scores,” which are relative measures of cell type abundance, not absolute cell counts. These scores indicate the degree to which a cell type’s gene signature is enriched within a sample compared to others in the analysis. A higher score suggests a greater presence of that cell type but does not provide a precise number of cells.
These scores are predictions that require careful interpretation within the biological context of the study. The accuracy depends on the quality of the input data and the relevance of the reference signatures to the tissue being studied. Therefore, xCell findings are a starting point for generating hypotheses that need further experimental confirmation.
In a research setting, validating these computational predictions is standard practice. Scientists use complementary laboratory techniques on a subset of samples to verify that xCell scores correlate with physical cell measurements. Common methods include flow cytometry, which sorts and counts individual cells, and immunohistochemistry (IHC), which visualizes specific cell types within a tissue slice.
For example, a study might use xCell to predict high levels of M1 macrophages in a set of tumor samples. To validate this, researchers would take sections from those same tumors and use IHC with an antibody specific to an M1 macrophage marker. If the IHC staining shows a high density of these cells, it provides experimental support for the xCell prediction.