Liquid biopsies are increasingly important diagnostic tools for early cancer detection due to their high sensitivity and minimal invasiveness, with cell-free RNA (cfRNA) emerging as a promising biomarker. At Flomics Biotech we developed a high-quality cfRNA-Seq platform that profiles human plasma cfRNA in a robust and reproducible manner. Our platform combines optimized laboratory protocols for Next-generation Sequencing of plasma RNA with state-of-the-art bioinformatics and machine learning (ML) classification methods to identify cancer-related RNA biomarker signatures. Here we used our cfRNA-Seq platform in the LiquiDx pre-clinical study with the aim of developing a multi-cancer early detection test.

We used our cfRNA-seq platform to profile the cfRNA in plasma samples from over 1,000 individuals (median age = 64 years) recruited in collaboration with 9 international clinical partners. The cohort consists of patients with colorectal, lung, breast, pancreatic or prostate cancer, or non-cancer diseases of the same organs, and healthy individuals. Early and late stage patients were recruited for each cancer type, with 22% of patients having stage I cancer. We isolated total RNA from 1 ml of plasma and sequenced at an average depth of 36M reads per sample with Illumina technology. Using the cfRNA-seq data we identified differentially expressed genes (DEGs), performed gene set enrichment analysis (GSEA), and trained an ML classifier to predict patient status and cancer tissue of origin. We employed a stratified 4-fold cross-validation strategy that maintains class balance across training and validation splits, and optimized the model’s hyperparameters.

We identified 114 DEGs (|log2FC| > 1, FDR < 0.05), with GSEA revealing enrichment of metastasis, inflammation, and proliferation-associated genes in cancer patient cfRNA profiles. For cancer vs healthy classification, our ML classifier achieves a mean area under the ROC curve of 0.92 ± 0.01, and a sensitivity of 83% at 90% specificity. For classification of different cancer types, at 90% specificity our ML classifier achieves a sensitivity ranging from 69% for breast cancer to 99% for prostate cancer. In contrast to other liquid biopsy-based technologies, our platform maintains a high performance level in detecting stage I cancer patients (80% sensitivity), reinforcing its great potential as a cancer early detection tool.

The Flomics cfRNA-Seq platform delivers high-quality cfRNA data and robust biomarker identification, demonstrating transformative potential in cancer early detection across multiple cancer types and stages, particularly in stage I where cancer detection is more challenging. The platform’s high performance and minimal invasiveness position it as a revolutionary tool to improve global cancer diagnostics.