Atopic dermatitis (AD) is a skin disease heterogeneous both in terms of clinical manifestations and molecular profiles. It is increasingly recognized that AD is a systemic rather than a local disease and should be assessed in the context of whole-body level biology. In this study, we integrated RNA-seq data of both skin and PBMC along with clinical data from 115 AD patients and matched 14 healthy controls aiming to comprehensively capture the molecular signature associated with specific clinical presentation. Analysis of cross-tissue ligand-receptor coupling suggested increase of skin-PBMC interactions in AD patients compared to healthy controls. We built a regression model that predicts clinical phenotypes of AD using transcriptome modules identified from weighted gene co-expression network analysis of RNA-seq data in skin tissue and PBMC. We identified differential immunological signatures associated with two qualitatively differential skin manifestations of AD, erythema and papulation. Furthermore, we applied the regression model established in the cross-sectional dataset to a longitudinal dataset collected monthly from 30 AD patients for up to a year for personalized monitoring of disease trajectory and examined longitudinal heterogeneity in association with clinical presentation. Three patient clusters identified on the basis of longitudinal features of blood tests and PBMC transcriptome modules were found to be associated with longitudinal features in clinical severity as well as in the medication history. Our approach serves as a framework for effective clinical investigation to gain a holistic view of pathophysiology of complex human diseases by highlighting inter- and intra- patient heterogeneity.