Inflammation is an essential response to disease, carrying both beneficial (elimination of pathogens) and detrimental (autoimmunity, cancer) effects. Inflammatory processes are carried out by a wide range of immune cell types (e.g. macrophages, T-cells, B-cells), which release and react to pro- and anti-inflammatory cytokines, and mediate immunological responses.
Quantification of immune cell subsets is essential to identify key cells involved in inflammation that can be used as prognostic biomarkers or drug targets, and be a crucial part of the underlying biological mechanism. For instance, high levels of circulating monocytes have been shown to be predictive of immunotherapy response in cancer patients, whereas low levels of CD4+ T-cells have been long defined as a hallmark of HIV infection. Single-cell technologies (e.g. flow-cytometry, scRNA-seq) have been used to quantify both circulating and tissue infiltrating immune cells from human samples. However, the total number of human samples with single-cell data is still very small compared to the existing body of bulk-level data (e.g. microarrays). This is statistically important, as adequate power and real-world biological and technical heterogeneity are needed to identify reproducible biological signals.
Cell-mixture deconvolution is a computational technique that complements single-cell experimental approaches by mapping patterns of bulk-level high-throughput data (e.g. RNA levels, DNA methylation) to estimated levels of immune cells, being effectively in silico flow-cytometry. Deconvolution is widely used in immunology to profile both blood and tissue biopsy bulk expression data from patients with inflammatory disease to identify immune cell biomarkers that can be diagnostic, prognostic, or therapeutic targets.
Deconvolution has the potential of extracting hidden biological information from vast amounts of existing bulk data, resulting in large gains in statistical power. However, deconvolution carries limitations that prevent current approaches from being consistently accurate across all datasets.
First, the accuracy of most methods is affected by both the technology used to profile the sample and its disease status. Second, most approaches focus on a specific data type (e.g. expression), neglecting nor integrating other sources (e.g. DNA methylation, cell-free DNA). Third, most approaches rely on a predefined set of cell types that are expected to be found in the samples of interest, preventing the discovery of new cell types from the data (e.g. new immune cell subtypes). The application of these new methods to existing data has the great potential of revealing new immunological knowledge and underlying mechanisms behind disease etiology and progression.
Recent work has has started to appreciate and tackle some of these issues, allowing new immunological discoveries by re-analyzing existing bulk level data. For example, deconvolution of public expression profiles from human cohorts revealed immune cells predictive of disease progression across different conditions, such as NK cells in Tuberculosis and monocytes in pulmonary fibrosis. These results highlight the importance of deconvolution in understanding the cellular basis of inflammation in human samples.
The scope of this topic is to push the boundaries of cell-mixture deconvolution and its application to inflammatory disease samples. These themes extend to but are not limited to:
-Improvements to reduce dataset confounders
-Multi ‘omics approaches and data integration
-Quantification of rare subsets (e.g. Th17, MDS cells)
-Discovery of unknown/unaccounted cell types
-Quantification of tissue resident immune cells (e.g. Langerhans cells, MAST cells)
-Measuring cell-type specific expression
-Deconvolution of cell-free DNA
Topic Editor Francesco Vallania is employed by Freenome Inc. All Topic Editors declare no competing interests with regards to the Research Topic subject.
Inflammation is an essential response to disease, carrying both beneficial (elimination of pathogens) and detrimental (autoimmunity, cancer) effects. Inflammatory processes are carried out by a wide range of immune cell types (e.g. macrophages, T-cells, B-cells), which release and react to pro- and anti-inflammatory cytokines, and mediate immunological responses.
Quantification of immune cell subsets is essential to identify key cells involved in inflammation that can be used as prognostic biomarkers or drug targets, and be a crucial part of the underlying biological mechanism. For instance, high levels of circulating monocytes have been shown to be predictive of immunotherapy response in cancer patients, whereas low levels of CD4+ T-cells have been long defined as a hallmark of HIV infection. Single-cell technologies (e.g. flow-cytometry, scRNA-seq) have been used to quantify both circulating and tissue infiltrating immune cells from human samples. However, the total number of human samples with single-cell data is still very small compared to the existing body of bulk-level data (e.g. microarrays). This is statistically important, as adequate power and real-world biological and technical heterogeneity are needed to identify reproducible biological signals.
Cell-mixture deconvolution is a computational technique that complements single-cell experimental approaches by mapping patterns of bulk-level high-throughput data (e.g. RNA levels, DNA methylation) to estimated levels of immune cells, being effectively in silico flow-cytometry. Deconvolution is widely used in immunology to profile both blood and tissue biopsy bulk expression data from patients with inflammatory disease to identify immune cell biomarkers that can be diagnostic, prognostic, or therapeutic targets.
Deconvolution has the potential of extracting hidden biological information from vast amounts of existing bulk data, resulting in large gains in statistical power. However, deconvolution carries limitations that prevent current approaches from being consistently accurate across all datasets.
First, the accuracy of most methods is affected by both the technology used to profile the sample and its disease status. Second, most approaches focus on a specific data type (e.g. expression), neglecting nor integrating other sources (e.g. DNA methylation, cell-free DNA). Third, most approaches rely on a predefined set of cell types that are expected to be found in the samples of interest, preventing the discovery of new cell types from the data (e.g. new immune cell subtypes). The application of these new methods to existing data has the great potential of revealing new immunological knowledge and underlying mechanisms behind disease etiology and progression.
Recent work has has started to appreciate and tackle some of these issues, allowing new immunological discoveries by re-analyzing existing bulk level data. For example, deconvolution of public expression profiles from human cohorts revealed immune cells predictive of disease progression across different conditions, such as NK cells in Tuberculosis and monocytes in pulmonary fibrosis. These results highlight the importance of deconvolution in understanding the cellular basis of inflammation in human samples.
The scope of this topic is to push the boundaries of cell-mixture deconvolution and its application to inflammatory disease samples. These themes extend to but are not limited to:
-Improvements to reduce dataset confounders
-Multi ‘omics approaches and data integration
-Quantification of rare subsets (e.g. Th17, MDS cells)
-Discovery of unknown/unaccounted cell types
-Quantification of tissue resident immune cells (e.g. Langerhans cells, MAST cells)
-Measuring cell-type specific expression
-Deconvolution of cell-free DNA
Topic Editor Francesco Vallania is employed by Freenome Inc. All Topic Editors declare no competing interests with regards to the Research Topic subject.
