The Impact of Leukocyte Retention on Platelet RNA-Based Cancer Detection

Platelet RNA profiling has emerged as a valuable modality in liquid biopsy, providing a minimally invasive window into host responses during systemic disease processes, including cancer. However, the impact of leukocyte retention within sample preparations on platelet transcriptomic signatures remains poorly understood. This study presents the first comprehensive assessment of how leukocyte retention influences platelet RNA sequencing data and, consequently, the performance of machine learning (ML)-based cancer detection. Using the largest publicly available platelet transcriptomic dataset to date, comprising 2,351 unique human RNA samples, supported by single-cell and single-platelet RNA sequencing from 29 independent donors, we quantified leukocyte retention across matched sample groups (low- and high- leukocyte retention) using a novel 138-gene leukocyte marker panel. To assess diagnostic implications, we employed 10 different ML algorithms in a binary classification task (cancer vs. healthy), and investigated the extent to which leukocyte-associated transcriptomic signals influence performance. ML models trained on high-retention samples often achieved slightly better classification performance, suggesting that leukocyte-derived transcripts may contribute complementary diagnostic information. Importantly, post hoc bioinformatic correction using remove unwanted variation (RUV) and linear modeling proved to effectively reduce the retention. These findings highlight the importance of controlling for leukocyte retention in platelet transcriptomic studies.