Atopic dermatitis is a common chronic disease that affects the quality of life of patients and their families and is marked by cyclical periods of flare-ups and remissions. Although there is no cure for the disease, its symptoms can be alleviated with proper management. However, atopic dermatitis is a complex disease that manifests differently in men and women, and the molecular mechanisms behind flare-ups are not yet fully understood. Identifying the genes that characterise lesional and non-lesional skin could improve the current understanding of the key molecular mechanisms and facilitate the development of new targeted therapies. In this paper, we present a machine learning approach aimed at discovering candidate sex-specific biomarkers for the disease. First, we selected the differentially expressed genes and applied feature selection techniques and machine learning methods to reduce the number of features. After a backward feature elimination step, we obtained an 11-gene male signature, a 10-gene female signature and an 8-gene general signature. Based on an independent test and using a soft-voting classifier, we obtained an AUC of 0.839 and accuracy of 0.7222 for the male signature and an AUC of 0.650 and accuracy of 0.6667 for the female signature. For the general signature, the AUC and accuracy values obtained were 0.783 and 0.8, respectively. The results suggest the potential existence of sex-specific biomarkers for atopic dermatitis. Consequently, our proposed gene signatures could serve as a starting point for scientific investigations aimed at understanding how the molecular mechanisms of the disease differ between males and females.