Enabling multiomic analysis in cell-free DNA tubes by addressing the impact of pre-analytical factors

Background : Cell-free DNA (cfDNA) tubes are widely adopted for genomic liquid biopsies because they allow delayed processing after blood collection while ensuring cfDNA stability. However, it remains poorly understood whether other analytes, especially plasma proteins, are stable in cfDNA tubes when subjected to preanalytical stresses during delayed processing. This study investigates the effect of time and temperature variations during transit on protein profiles in samples collected in Streck cfDNA tubes, and a computational method to mitigate the impact. Methods: We collected blood from 10 healthy donors into Streck cfDNA tubes. Plasma was separated at varying time points and after exposure to different temperature conditions to simulate shipping stress. A total of 2,903 proteins were measured using an immuno-proteomic profiling platform to assess changes in protein abundance. We also assessed a computational regression approach to correct for preanalytical impact on protein levels measured by a bead-based multiplex immunoassay and determine the effect size change of biomarker candidates in a clinical cohort (cancer n=105 vs. control n=538). Results: Using a 20% abundance change cutoff, we observed that over half of the proteins (1,574) remained stable over prolonged time and temperature exposure. While 351 proteins showed negative bias, the majority of perturbations (978) were concentration increases, with proteins from blood cells overrepresented. Some proteins, such as Macrophage Migration Inhibitory Factor (MIF), were highly correlated with plasma hemoglobin (Hb) level, indicating a red blood cell (RBC) lysis effect. Other proteins, such as Epidermal Growth Factor (EGF), were more correlated with plasma potassium (K) level, suggesting broader lysis effects involving other blood cell types. Using the independent clinical cohort, we confirmed that MIF and EGF levels were highly correlated with Hb and K, respectively. The regression results showed that the Hedges' g effect size for MIF between case and control groups increased from 0.184 to 0.468 after Hb-regression, and the EGF Hedges' g increased from 0.318 to 0.467 after K-regression, demonstrating the power of this computational method to unmask potential protein biomarkers. Conclusions: Preanalytical variables significantly impact plasma proteomic profiles in cfDNA tubes, especially for proteins that are associated with blood cells. The protein abundance changes in response to preanalytical stresses in cfDNA tubes can confound analyses and lead to false findings. Our results demonstrate that leveraging preanalytical markers like K and Hb allows for computational mitigation of this variability. Implementing robust sample quality control and computational correction strategies is essential to ensure the reliability of plasma proteomic measurements from samples collected in cfDNA tubes.