Tumor cellular composition—including malignant cell states, immune populations, and stromal cells—is increasingly recognized as a determinant of therapeutic response and resistance. However, comprehensive single-cell level profiling has largely remained in research settings. This study describes a clinically compatible sample-to-report workflow that performs tumor composition profiling from routine FFPE (formalin-fixed paraffin-embedded) clinical specimens, enabling prospective, sample-by-sample analysis without requiring dedicated research material or cohort-based processing.
The workflow integrates low-input single-nucleus RNA sequencing with foundation model–based automated cell annotation. By adapting single-nucleus methods to routine FFPE sections, the pipeline supports analysis of nuclei derived from standard pathology blocks. Automated cell annotation leverages foundation model techniques to assign cell identities at scale, producing cell-type labels and cell-type–specific gene expression profiles suitable for a clinical reporting context.
Across a clinical cohort of 116 specimens drawn from routine pathology, the authors applied the workflow to six distinct human cancer types. The study expressly demonstrates that the method is compatible with the types of specimens routinely available in clinical practice. The sample set included both contemporary clinical FFPE sections and archival material, and the approach was also tested on ultra-low-input biopsy specimens to assess performance on very small samples.
Application of the workflow produced reproducible measurements of overall cellular composition and of cell-type–specific gene expression across the cohort. The method generated data at the resolution required to distinguish malignant cell states, immune subpopulations, and stromal elements, enabling quantification of the cellular constituents that compose the tumor microenvironment. These measurements were consistent across samples processed with the described pipeline, supporting its use for routine profiling.
The authors report high technical reproducibility for the single-nucleus transcriptomic measurements derived from routine FFPE sections. Importantly, cellular composition estimates from the single-nucleus data showed concordance with traditional pathological assessments of immune infiltration, indicating that the molecular readouts align with established histopathological evaluation. This concordance supports the potential clinical validity of the workflow when comparing molecular cell-type estimates to pathologist-derived metrics.
Beyond freshly processed routine sections, the workflow was found to be applicable to archival FFPE material, extending its potential to historical pathology collections. The authors also demonstrated feasibility on ultra-low-input biopsy specimens, indicating the method can be applied to small diagnostic samples commonly obtained in the clinic. These capabilities suggest broad applicability across the lifecycle of clinical sample availability, from prospective diagnostics to retrospective research or biomarker validation using archived tissues.
By enabling single-cell profiling directly from standard clinical pathology specimens, this workflow establishes a practical framework for routine molecular characterization of the tumor microenvironment. The ability to prospectively measure cellular composition sample by sample opens the possibility of evaluating cellular metrics as clinical biomarkers in precision oncology. Routine application could inform therapeutic decision-making by providing detailed immune and stromal context alongside genomic or histopathological information.
This report is presented as a preprint; the paper notes competing interests and intellectual property activity tied to the work. Specifically, Celine Vallot is disclosed as a co-founder of One Biosciences, and patents have been filed related to the reported methods. The preprint documents the workflow and cohort-level results but does not replace prospective clinical validation studies required to establish clinical utility. Further work will be needed to define analytic performance characteristics in regulated settings, to validate biomarker associations with clinical outcomes, and to standardize reporting and integration with clinical workflows.
The study demonstrates that routine FFPE sections can support clinically compatible single-nucleus RNA sequencing to profile tumor cellular composition across multiple cancer types. The approach is reproducible, concordant with pathological assessment of immune infiltration, and applicable to archival and ultra-low-input specimens. These features make it a promising framework for integrating single-cell–level cellular composition profiling into clinical oncology practice, pending additional validation and regulatory steps.