A multivariate cell-based liquid biopsy for lung nodule risk stratification: Analytical validation and early clinical...

The Indicator Cell Assay Platform (iCAP) is a novel tool for blood-based diagnostics that uses living cells as biosensors to integrate and amplify weak, multivalent disease signals present in patient serum. In the platform, standardized cells are exposed to small volumes of patient serum, and the resulting transcriptomic response is analyzed using machine learning tools to develop disease classifiers.

We developed a lung cancer-specific iCAP (LC-iCAP) as a rule-out test for the management of indeterminate pulmonary nodules detected by low-dose CT screening. This included assay parameterization, analytical reproducibility testing, and selection of a fixed 85-gene feature set for future clinical validation and regulatory development. Clinical performance was estimated using a prospective-specimen-collection, retrospective-blinded-evaluation (PRoBE) study design comprising 176 samples. Classifier variants were trained by nested cross validation using subsets of the 85 genes, and selected variants were evaluated by temporal blind validation using 39 control and 40 case samples (72 % Stage I, 22 % Stage II cancer).

The assay showed excellent reproducibility across various conditions and cell lineages, and case versus control transcriptomic signals were enriched for hypoxia-responsive genes, consistent with known lung cancer biology. Two models demonstrated discriminative ability in blind validation, one with AUC = 0.64 (95 % CI: 0.51-0.76). Post hoc integration with CT imaging features yielded a combined model with 90 % sensitivity, 64 % specificity, and 95 % negative predictive value at 25 % prevalence, suggesting clinical utility and surpassing performance of existing rule-out tests.

This study establishes the analytical reproducibility and biological relevance of the LC-iCAP. While clinical validation is preliminary, the results support the assay's potential utility in lung nodule management. The study introduces a new paradigm of using scalable and cost-effective cell-based biosensor assays for liquid biopsies. With a multivariate readout, the platform is amenable to precision medicine applications such as multi-cancer early detection.