Detección directa de células tumorales raras circulantes en células mononucleares de sangre periférica mediante scRNA...

Circulating tumor cells (CTC) provide a minimally invasive window into metastatic disease but are difficult to detect and estimate due to their rarity and heterogeneity. Conventional enrichment-based approaches introduce selection bias and fail to capture diverse CTC populations. Single-cell RNA sequencing (scRNA seq) enables unbiased transcriptomic profiling of diverse cell types and rare population within complex samples like blood. Here, we evaluated feasibility of a computational spike-in framework to assess the sensitivity and specificity of scRNA seq based CTC detection in a peripheral blood background.

Three PBMC datasets comprising 20,871; 14,367; and 13,731 cells were created by merging Cell Ranger-derived raw matrices from 12 PBMC samples (4 per dataset). Cervical cancer (CaCx) dataset (SRR13927092) raw matrices were prepared similarly. CaCx cells were randomly selected and spiked at levels of 50, 25, 10, 5, and 2 cells into each dataset, with three replicates per level. Linear regression, limit of detection (LOD) and limit of quantification (LOQ) estimation, were performed.

Unsupervised gene expression profiling revealed distinct clusters of CaCx cells in PBMCs background using k-means clustering. Clustering with k-mean value 7 resulted specific CaCx clusters. The average detection efficiency ranged from 66% to 93% for unsupervised clustering. Supervised clustering with specific epithelial markers improved identification, achieving 95%-100% detection accuracy. Linear regression showed a high coefficient of determination (R² = 0.9991). The estimated LOD was 1.0 cell, while LOQ was around 3.3 cells.

This study confirms that single-cell analysis pipelines are competent, can effectively and correctly detect rare epithelial tumor cells in PBMCs with high sensitivity and reproducibility, even at very low concentrations.