Beyond tumor-shed markers: AI driven tumor-educated polymorphonuclear granulocytes monitoring for multi-cancer early detection.

Abstract

3060 Background: Tumor educated Polymorphonuclear Granulocytes (tPMNG) are a distinct phenotypic set of Neutrophils (N2) with corrupted programmed death pathways resulting in apoptosis resistance. The presence of tPMNGs in the peripheral blood indicates up-regulation of pro-tumoral factors and resistance to apoptotic signals. Our platform leverages this unique anti-apoptotic characteristic of tPMNGs through a proprietary culture process and AI-based digital imaging analysis to detect cancer in its early stages. This approach represents a paradigm shift from detecting rare tumor-shed analytes such as ctDNA and CTCs to monitoring relatively abundant tPMNGs, whose numbers typically exceed conventional analytes by 3-4 orders of magnitude. We studied detection of tPMNGs in a case-control study to evaluate their suitability for a multi-cancer early detection test (MCED). Methods: We collected 10 ml of peripheral blood in EDTA tubes from 892 asymptomatic healthy volunteers above 18 years of age [463, 52% male; 429, 48% females with mean age of 48 (20 to 89) years], 24 individuals diagnosed with non-malignant conditions including prostatitis, polycystic ovarian disease and acute pancreatitis, and 90 individuals recently diagnosed with surgically resectable early stage cancers (Stage 1/ 2) comprising Head and Neck (N=32, 36%), Breast (N=20, 22%), Colorectal (N=14, 16%), others (N=24, 27%). Nucleated cells were isolated from the samples after RBC lysis and centrifugation. These cells were seeded in six well culture plates and subjected to controlled apoptotic stress under serum-free, hypoxic conditions with specific growth factor supplementation for 5 days. Surviving cells were set on imaging slides and stained with H&E. 60X images were obtained and analyzed using a convolutional neural network (CNN) based AI algorithm to detect tPMNGs per ml. Results: The mPMNG detection method demonstrated 84% sensitivity (95% CI: 84.44%) across multiple cancer types. The platform demonstrated 97% specificity (95% CI: 97.40%) among healthy asymptomatic cohort. In samples from individuals from individuals with non-malignant conditions, specificity was 96% (95% CI: 95.83%). Conclusions: This first-in-class immune cell-based MCED approach offers several advantages over tumor-shed analyte detection: 1. Leverages amplified host response rather than rare tumor products. 2. Provides robust detection across cancer types and stages. 3. Utilizes existing laboratory infrastructure and a scalable protocol. 4. Demonstrates potential for screening, diagnosis, and monitoring applications. The high sensitivity and specificity, combined with practical advantages, suggest potential for clinical implementation in cancer screening and monitoring programs either using tPMNG alone or in conjunction with CTCs / cfDNA evaluation.