The aim was to identify genetic variants associated with refined asthma phenotypes enabling to take into account multiple features of the disease.Latent class analysis (LCA) was applied in 3001 adults ever having asthma recruited in the frame of three epidemiological surveys (ECRHS, SAPALDIA, EGEA). Fourteen personal and phenotypic characteristics from questionnaires and clinical examination were used. A genome wide association study (GWAS) was conducted for each LCA-derived asthma phenotype, compared to subjects without asthma (n=3474).The LCA identified four adult asthma phenotypes, mainly characterized by the activity of the disease, the age of asthma onset and the atopic status. Associations of genome wide significance (<1.25×10-7) were observed between "active adult-onset non-allergic asthma" and rs9851461 flanking CD200 (3q13.2) and between "Inactive/mild non-allergic asthma" and rs2579931 flanking GRIK2 (6q16.3). Borderline significant results (2.5×10(-7).