The heterogeneity of the schizophrenia phenotype is often considered an obstacle for genetic research. We therefore aim to identify subgroups of psychosis patients with a shared symptom profile by means of a fully data-driven analysis, which may serve as an alternative phenotype. We investigated the symptoms of 1056 patients that were referred to our hospital with a psychosis. The lifetime symptoms scores were derived from the current and lifetime ratings of the comprehensive assessment of psychiatric history (CASH) interview. We used latent class analysis (LCA) to identify clusters of patients with a shared symptom profile. The five indicators in our analysis were the total number of symptoms present for each of the five factors identified in a factor analysis of lifetime symptoms. We also analysed the discriminating power of these symptom dimensions in previous LCAs. A six-cluster division of psychotic phenotypes showed substantial overlap with earlier LCA analyses and findings from genetic association studies. The results included a bipolar and a depression subgroup in psychosis and showed that mood symptoms are the best discriminators of subgroups of psychosis. The distinction of subgroups of psychosis patients, in particular those with major mood symptoms could facilitate the unravelling of the genetics of psychotic disorders.