To evaluate whether endometrial molecular profiles distinguish subsets of patients according to clinical characteristics, and to infer dysregulated immune networks, by measuring cytokines, chemokines, and growth factors in endometrial biopsy specimens from a cohort of infertile women with a high incidence of endometriosis. Prospective cohort study. Department of Gynecology at a university hospital. Patients undergoing laparoscopy for infertility assessment (n = 103). Endometrial biopsies were performed during surgery. Fertility outcome and clinical parameters were registered preoperatively and after 6 months. The concentrations of 48 factors in endometrial biopsy specimens were analyzed with respect to clinical status in univariate and multivariate frameworks. The concentrations of 44 factors from endometrial tissues of 74 patients were suitable for analysis. Although the tissue concentrations of interleukin (IL)15, IL-7, and interferon γ-induced protein (IP)-10 were individually lower in patients with endometriosis than in those without endometriosis, the differences were not significant after multiple comparison. However, multivariate modeling incorporating covariation showed separation between subsets of endometriotic and nonendometriotic patients, based predominantly on IP-10, monocyte chemoattractant protein-1, IL-16, and IL-18; this result was independent of cycle and fertility status. Analysis restricted to endometrial tissues from the secretory phase separated endometriotic and nonendometriotic patients by a combination of IL-15, IP-10, monocyte chemoattractant protein-1, IL-16, and IL-18. This combination suggests a uterine natural killer cell defect. We found no significant correlations between endometrial cytokines and fertility outcome. A molecular signature in endometrial tissue was able to distinguish endometriotic from nonendometriotic patients, implicating uterine natural killer cells in endometriosis.