Obese subjects with a similar body mass index (BMI) exhibit substantial heterogeneity in gluco-and cardiometabolic heath phenotypes. However, defining genes that underlie the heterogeneity of metabolic features among obese individuals and determining metabolically healthy and unhealthy phenotypes remain challenging. We conducted unsupervised hierarchical clustering analysis of subcutaneous adipose tissue transcripts from 30 obese men and women⩾ 40 years old. Despite similar BMIs in all subjects, we found two distinct subgroups, one metabolically healthy (group 1) and one metabolically unhealthy (group 2). Subjects in group 2 showed significantly higher total cholesterol (P= 0.005), low-density lipoprotein cholesterol (P= 0.006), 2-h insulin during oral glucose tolerance test (P= 0.015) and lower insulin sensitivity (S I, P= 0.029) compared with group 1. We identified significant upregulation of 141 genes (for example, MMP9 and SPP1) and downregulation of 17 genes (for example, NDRG4 and GINS3) in group 2 subjects. Intriguingly, these differentially expressed transcripts were enriched for genes involved in cardiovascular disease-related processes (P= 2.81× 10− 11–3.74× 10− 02) and pathways involved in immune and inflammatory response (P= 8.32× 10− 5–0.04). Two downregulated genes, NDRG4 and GINS3, have been located in a genomic interval associated with cardiac repolarization in published GWASs and zebra fish knockout models. Our study provides evidence that perturbations in the adipose tissue gene expression network are important in defining metabolic health in obese subjects.