Cell-free RNA (cfRNA) circulating in blood plasma can be decomposed into cell-type-specific contributions only if a reference 'basis matrix' of gene-expression profiles covering every contributing cell type exists. Vorperian et al. (Nature Biotechnology, 2022) demonstrated such cell-type decomposition of the plasma cell-free transcriptome using the Tabula Sapiens whole-body single-cell atlas, but the study's central unresolved issue is that 'a complete reference dataset spanning the entire cell type space of the human body does not yet exist': brain and testis cell types are absent from whole-body atlases and required a separate, less-integrated signature-scoring approach; transcriptionally similar cell types shared across tissues had to be grouped rather than resolved; and inconsistent cell-type annotations across independent atlases corrupt the definition of cell-type-specific genes. The open problem is to construct a complete, annotation-harmonized single-cell reference basis and integrate it into a single validated deconvolution of plasma cfRNA. Benchmarking in March 2026 reports that cell-type-level inference remains inconsistent across methods and references, with missing brain cell types in standard references causing systematic misassignment of cfRNA signals.

Background

Cell-free RNA (cfRNA) is RNA released into blood plasma by cells throughout the body — passively, during cell death (apoptosis or necrosis), and/or actively, for example packaged in extracellular vesicles (lipid-bound particles secreted by living cells). Because plasma cfRNA can be collected by a simple blood draw and sequenced, it serves as a 'liquid biopsy': a noninvasive readout of physiological and disease states. Existing clinical applications — fetal and placental monitoring in pregnancy, and organ-injury signatures in transplant, liver, kidney and neurodegenerative disease — mostly infer the tissue of origin of cfRNA molecules, yet the pathophysiology being measured operates at the level of cell types.

Deconvolution is the standard computational route to finer resolution: the measured cfRNA expression profile is modeled as a mixture x≈∑kckbkx \approx \sum_k c_k b_k, where each column bkb_k of a 'basis matrix' is a reference gene-expression profile for one cell type and ck≥0c_k \ge 0 is that cell type's fractional contribution. Vorperian and colleagues (Nature Biotechnology, 2022) applied this idea body-wide using nu-SVR (nu-support vector regression, a regression variant previously used to decompose bulk tissue transcriptomes into cell-type fractions) with a basis matrix built from Tabula Sapiens, a multiple-donor whole-body single-cell RNA-sequencing atlas spanning 24 organs (Science, 2022). They showed that platelet, erythroid and leukocyte signals dominate the healthy plasma cell-free transcriptome; that contributions from solid-organ cell types (intestine, liver, lung, pancreas, heart, kidney) are simultaneously resolvable; and that cell-type-specific pathology is noninvasively measurable — for example, loss of proximal-tubule signal in chronic kidney disease and reduced neuronal and glial signals in Alzheimer's disease.

The 2022 study, however, depends on reference data that do not exist in complete form. As the authors state, 'a complete reference dataset spanning the entire cell type space of the human body does not yet exist.' Tabula Sapiens omits, among others, brain and testis, so the authors worked around the gap by intersecting cfRNA genes with the Human Protein Atlas RNA consensus dataset (a curated body-wide RNA resource) and by scoring signatures defined from individual organ atlases — for example an Alzheimer's-disease brain single-cell dataset, a human liver cell atlas and placenta single-cell atlases — a separate scoring step that is less integrated than the simultaneous deconvolution because it does not consider all cell types at once. Three further consequences of reference incompleteness appear in the source: transcriptionally similar cell types occurring in several tissues (for example, vascular endothelial cells of different organs) had to be grouped to control multicollinearity (near-linear dependence among reference profiles that makes individual contributions unidentifiable), so cross-tissue cell-type resolution was abandoned rather than achieved; annotation discrepancies between independent single-cell atlases of the same tissue (illustrated with placental trophoblast subtypes, where an earlier cfRNA study had used genes that are not specific to their asserted cell type) propagate into which genes are called cell-type-specific; and the framework was restricted to cell types unique to individual tissues.

The missing reference is the goal of the Human Cell Atlas program (launched 2017), which aims to build comprehensive reference maps of all human cell types, but as of the end of 2025 no complete atlas exists: a comparative review of the single-cell atlas landscape still reports critical gaps and underrepresented tissue types, organ atlases are accumulating piecemeal, and whole-body multi-donor atlases such as Tabula Sapiens still exclude brain and testis. For liquid biopsy this is not an abstract inconvenience: systematic benchmarking in 2026 shows that the missing brain cell types in standard references cause misassignment of cfRNA signals (Schwann-cell profiles acting as surrogates for absent neuronal populations), and that cell-type-level deconvolution of cfRNA remains far less reliable than tissue-level inference.

Problem Statement

Remove the reference-completeness bottleneck that limits cell-type decomposition of the human plasma cell-free transcriptome, as posed by Vorperian et al. (Nature Biotechnology, 2022): 'a complete reference dataset spanning the entire cell type space of the human body does not yet exist.' Construct a single-cell reference basis that spans the entire human cell-type space — including cell types absent from current whole-body atlases such as brain and testis, and transcriptionally similar cell types shared across tissues (for example, organ-specific vascular endothelial cells and fibroblasts, which current pipelines must group) — with cell-type annotations harmonized across source atlases so that the definition of cell-type-specific genes does not depend on which atlas is consulted; integrate this basis into a single simultaneous deconvolution of plasma cell-free RNA (cfRNA), with no separate signature-scoring fallback for uncovered cell types; and validate that the inferred cell-type contributions are accurate at the cell-type level, using ground truth from experimental cfRNA mixtures and/or clinical cohorts with histologically confirmed cell-type-specific injury. Because the cell-type space is set by evolving single-cell inventories, the deliverable must fix its own boundaries: state the pinned cell-type inventory the basis is complete against (a named cell-type ontology or an explicitly enumerated cell-type list that includes the brain and testis cell types absent from whole-body atlases) and report coverage against it, and state the explicit acceptance criterion for cell-type-level accuracy (a metric with a threshold, or the named qualifying ground-truth datasets) together with the accuracy obtained against it.

References

  1. Vorperian, S. K., Moufarrej, M. N., Tabula Sapiens Consortium, Jones, R. C., Karkanias, J., Krasnow, M., Pisco, A. O., Quake, S. R., Salzman, J., Yosef, N., et al. Cell types of origin of the cell-free transcriptome. Nature Biotechnology 40, 855-861 (2022). DOI 10.1038/s41587-021-01188-9. https://doi.org/10.1038/s41587-021-01188-9 (open-access full text, PMC9200634; the peer-reviewed version of the source paper)
  2. Vorperian, S. K., Moufarrej, M. N., Tabula Sapiens Consortium, Quake, S. R. Cell Types of Origin in the Cell Free Transcriptome in Human Health and Disease. bioRxiv preprint, posted 2021-05-05. DOI 10.1101/2021.05.05.441859. https://doi.org/10.1101/2021.05.05.441859 (preprint of the Nature Biotechnology article; the nine-item limitation list is this version's Discussion)
  3. The Tabula Sapiens Consortium, Jones, R. C., Karkanias, J., Krasnow, M. A., et al. The Tabula Sapiens: A multiple-organ, single-cell transcriptomic atlas of humans. Science 376, eabl4896 (2022). DOI 10.1126/science.abl4896. https://doi.org/10.1126/science.abl4896
  4. Regev, A., Teichmann, S. A., Lander, E. S., et al. The Human Cell Atlas. eLife 6, e27041 (2017). DOI 10.7554/eLife.27041. https://doi.org/10.7554/eLife.27041
  5. Uhlen, M., Karlsson, M. J., Zhong, W., et al. A genome-wide transcriptomic analysis of protein-coding genes in human blood cells. Science 366, eaax9198 (2019). DOI 10.1126/science.aax9198. https://doi.org/10.1126/science.aax9198 (Human Protein Atlas RNA consensus dataset)
  6. Mathys, H., Davila-Velderrain, J., Peng, Z., et al. Single-cell transcriptomic analysis of Alzheimer's disease. Nature 570, 332-337 (2019). DOI 10.1038/s41586-019-1195-2. https://doi.org/10.1038/s41586-019-1195-2 (brain cell-type source used to work around atlas absence)
  7. Aizarani, N., Saviano, A., et al. A human liver cell atlas reveals heterogeneity and epithelial progenitors. Nature 572, 199-204 (2019). DOI 10.1038/s41586-019-1373-2. https://doi.org/10.1038/s41586-019-1373-2 (organ atlas used by the source)
  8. Vento-Tormo, R., Efremova, M., Botting, R. A., et al. Single-cell reconstruction of the early maternal-fetal interface in humans. Nature 563, 347-353 (2018). DOI 10.1038/s41586-018-0698-6. https://doi.org/10.1038/s41586-018-0698-6 (placenta atlas illustrating cross-atlas annotation discrepancy)
  9. Tsang, J. C. H., Vong, J. S. L., Ji, L., et al. Integrative single-cell and cell-free plasma RNA transcriptomics elucidates placental cellular dynamics. Proceedings of the National Academy of Sciences 114, E7786-E7795 (2017). DOI 10.1073/pnas.1710470114. https://doi.org/10.1073/pnas.1710470114 (earlier trophoblast cfRNA study whose gene specificity the source questioned)
  10. Vorperian, S. K., Dennis, L. M., Hupalowska, A., Rood, J. E., Quake, S. R. Cell type inference in cell-free nucleic acid liquid biopsy. Nature Biotechnology 43, 1940-1953 (2025). DOI 10.1038/s41587-025-02904-5. https://doi.org/10.1038/s41587-025-02904-5 (review by the source's authors; highlights outstanding questions)
  11. Ioannou, A., Friman, E. T., Daub, C. O., Bickmore, W. A., Biddle, S. C. Benchmarking tissue- and cell type-of-origin deconvolution in cell-free transcriptomics. bioRxiv preprint, posted 2026-03-09. DOI 10.64898/2026.03.05.709833. https://doi.org/10.64898/2026.03.05.709833
  12. Yang, T., Qin, Y., Yan, S., Guo, S., Sun, J., Huang, J., Li, J., Zhou, Q., Jin, X., Wang, W.-J. Comprehensive evaluation of methods for identifying tissues or cell types of origin of the plasma cell-free transcriptome. PeerJ 13, e19241 (2025). DOI 10.7717/peerj.19241. https://doi.org/10.7717/peerj.19241
  13. Yan, S., Tian, X., Qin, Y., Li, J., Yang, T., Yu, D., Liu, W., Sun, J., Hu, C., Zhou, Q., Liu, Z., Wang, W.-J. Pathway-enhanced Transformer-based robust model for quantifying cell types of origin of cell-free transcriptome (Deconformer). bioRxiv preprint, posted 2024-02-28. DOI 10.1101/2024.02.28.582494. https://doi.org/10.1101/2024.02.28.582494
  14. Look, K., Goh, P. E., Ming, L. C., Law, S. H., Tan, S. L., Tan, E. W., Ser, H. L., Goh, B. H. A comparative review of single-cell atlases: mapping cellular diversity across species and tissues. Cellular and Molecular Life Sciences 83, 22 (2025). DOI 10.1007/s00018-025-05952-x. https://doi.org/10.1007/s00018-025-05952-x