Background <p>Syndrome differentiation in Traditional Chinese Medicine (TCM) is pivotal to clinical practice and dictates the efficacy of medicinal treatments. However, precision diagnostic models for TCM syndromes, constructed from biomarkers such as metabolites and proteins, have failed to achieve high precision. Recent studies have highlighted a strong link between TCM and epigenetics, an area that remains largely unexplored in TCM diagnosis. Taking atrial fibrillation (AF) with Qi-Yin deficiency syndrome (QYDS) as an example, we utilized a type of epigenetic sequencing technology called 5hmC-Seal and integrated it with various machine learning models to develop an Epigenetic Differential Syndrome (Epi-DS) technology for identifying epigenetic biomarkers. This approach is crucial for developing more accurate diagnostic models for traditional Chinese medicine syndromes and for advancing the modernization of traditional Chinese medicine.</p> Methods <p>In this study, we conducted a single-center, prospective study involving two independent cohorts (cohort 1 and cohort 2) in AF, including QYDS and non-Qi-Yin deficiency syndrome (NQYDS). Next, we utilized 5hmC-Seal to obtain the patients’ 5hmC genome-wide profiles in plasma extracellular vesicles DNAs (evDNAs). Meanwhile, a variety of sophisticated machine learning algorithms were employed across three datasets—training, validation, and external cohorts (the training and validation sets constituting cohort 1 and the external cohort constituting cohort 2) to construct and validate QYDS diagnosis model.</p> Results <p>Based on the hydroxymethylation profile of the QYDS in AF, we have successfully constructed a disease-phenotype-molecule biological network for AF. At the molecular level, we identified nine characteristic 5hmC markers for the QYDS in AF and successfully established a diagnostic model for this syndrome. In Cohort 1’s training set, the area under the receiver operating characteristic curve (AUC) was as high as 0.984, with a sensitivity of 0.976 and a specificity of 1.000. In validation set, the AUC was 0.949, with a sensitivity of 0.952 and a specificity of 0.952. In the independent external validation cohort 2, the AUC was as high as 0.934, with a sensitivity of 0.886 and a specificity of 0.919. Moreover, the diagnostic model we built based on symptoms and molecular markers achieved an AUC value of 0.864 in an independent external cohort.</p> Conclusions <p>A novel precision diagnostic approach of TCM Syndrome Differentiation was established based on Epi-DS. The disease-phenotype-molecule network we have constructed reveals the epigenetic foundation of TCM and has identified molecular diagnostic markers for the QYDS in AF. This provides an example for understanding the molecular basis of TCM syndrome differentiation and for integrated traditional Chinese and Western medicine diagnosis.</p>