PATIENT STRATIFICATION USING LATENT VARIABLES

A computer-implemented method of stratifying a population of patients into disease endotypes is provided. The method comprises: encoding (202), using an unsupervised machine learning model, data (102, 400) relating to the patients as latent variables, wherein the latent variables represent different groupings of related biological features; determining (204) one or more importance measures of the latent variables, wherein the determining comprises encoding, using the unsupervised machine learning model, the data relating to the patients a plurality of times and determining an extent of recurrence of the latent variables; prioritising (206) the latent variables using the importance measures, the prioritising comprising ranking the latent variables from most important to least important for the disease; interpreting (208) one or more of the latent variables above a priority threshold, wherein the interpreting comprises identifying genes that the latent variable encodes and identifying pathological gene expression patterns that the latent variables represent by applying gene enrichment analysis (620); and identifying (210) a disease endotype that is represented by one or more of the interpreted latent variables, wherein the identifying comprises identifying a biological process underlying the disease using a gene expression pattern encoded in the one or more latent variables.