Abstract 2278: Correlation between DNA-methylation predicted and measured immune-cell type proportions by age in ARIC

Abstract Background: DNA methylation (DNAm)-predicted deconvolution uses cell-specific methylation markers of leukocytes in peripheral blood to estimate immune cell type proportions. Validity is high for predicted proportions of immune cell types. Given known age-related immune system changes and changes in predicted immune cell proportions by age, we assessed whether the correlations between predicted and measured proportions differ across age. Such information is needed for cancer studies on immunoaging in cohorts without CBC differential or stored whole blood. Methods: Lymphocyte, neutrophil, monocyte, eosinophil, and basophil proportions were previously estimated using DNAm deconvolution (Salas et al., PMID: 35140201) and measured by CBC among Black participants (N=151; 62% female) without cancer in the Atherosclerosis Risk in Communities study. Pearson correlation coefficients were calculated in age tertiles (T1: 47-53, T2: 54-60, T3: 61-67 years). To minimize confounding by health factors, we stratified by more (no diabetes, BMI <30 kg/m2, and non-smoker) and fewer (≤ 2 of the factors) positive health factors. Correlations in age T1 and T3 were compared after Fisher’s z-transformation. Results: Correlations between predicted and measured proportion were similar across age tertiles, albeit slightly weaker in the oldest tertile (Table, except for basophils). This same pattern was present stratifying by health factors, although the correlations appeared slightly stronger for more vs. fewer positive health factors (e.g., age T1, T3 - neutrophils: more 0.81, 0.77; fewer 0.78, 0.66). However, all p-values were > 0.1 comparing correlations in the youngest and oldest age tertiles. Conclusion: For each common immune cell type, correlations between DNAm-predicted and measured proportion were generally similar across mid- to older adulthood, supporting use of the algorithm in immunoaging studies. Studies with participant diversity, including ages ≥ 70 yrs are needed. Citation Format: Zhike Lin, Dominique S. Michaud, Naisi Zhao, Vernon A. Burk, Eric Boerwinkle, Jan Bressler, Devin C. Koestler, Karl T. Kelsey, Elizabeth A. Platz. Correlation between DNA-methylation predicted and measured immune-cell type proportions by age in ARIC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2278.