Understanding which cell types are present in a tissue is essential for interpreting gene expression data, especially when samples contain a mixture of many different cells. Bulk transcriptomics measures the average gene expression across all cells in a sample, which can hide important cell specific signals. To overcome this, scientists use cell type deconvolution methods that estimate the proportions of different cell types from bulk RNA data.

Researchers at the Technical University of Munich and the University of Innsbruck have developed and evaluated a comprehensive framework to assess modern deconvolution tools. These newer, second generation methods rely on single cell RNA sequencing data to build flexible reference profiles, allowing them to estimate a wide range of cell types across many tissues and organisms. While powerful, this flexibility also makes it harder to compare tools and understand when and why they succeed or fail.

omnideconv ecosystem and benchmark. A The omnideconv benchmarking ecosystem offers five tools (from left to right): the R package omnideconv providing a unified interface to deconvolution methods, the pseudo-bulk simulation method SimBu, the deconvData data repository, the deconvBench benchmarking pipeline in Nextflow, and the web-app deconvExplorer. B Outline of the benchmark experiment: scRNA-seq and bulk RNA-seq data are used as input for several methods, and a unified output of estimated cell-type fractions for each bulk sample is calculated. We compare the estimated fractions to ground-truth fractions (known from pseudo-bulks or FACS/IHC/CBC assays) and compute performance measures per method and cell type. C Several challenges in cell-type deconvolution are addressed in this benchmark: (1) mRNA content bias across different cell types; (2) a fraction of the cells may be of unknown type, since the scRNA-seq reference does not necessarily contain all cell types present in the bulk mixture; (3) the level of annotation granularity; (4) some cell types are more similar on a transcriptomic level, leading to “spillover” towards similar cell types; (5) scRNA-seq datasets vary by technology, tissue and batch effects.

The team systematically benchmarked many state of the art deconvolution methods using both real biological datasets and carefully designed simulations. By varying factors such as how similar cell types are to each other, how complete the reference data are, and where the datasets originated, they showed that performance can differ substantially between methods. Some tools were more accurate but slower, while others scaled well to large datasets but struggled when cell types were closely related.

Importantly, the researchers found that no single method was best in all situations. Instead, different tools showed strengths under different conditions, highlighting the need to match the deconvolution approach to the biological question and data type. To make this easier for the community, they introduced omnideconv, a unified ecosystem that brings together multiple deconvolution methods, benchmarking datasets, and evaluation workflows in one place.

By providing a standardized way to apply, compare, and optimize deconvolution tools, omnideconv helps researchers make more informed choices when analyzing bulk RNA sequencing data. This work supports more reliable interpretation of complex tissues and improves the integration of bulk and single cell transcriptomics in biomedical research.

Source – A Docker image that includes an installed version of omnideconv and its dependencies can be accessed via Dockerhub at https://hub.docker.com/repository/docker/alexd13/omnideconv_benchmark/general